22 | Competition gets heated with Deepseek vs. OpenAI

Kicking off with AI reflections reminiscent of Black Mirror, we dive into how emerging tools are breaking perception barriers. Chris shares exciting AI tools under development, while we also spotlight DeepSeek, a groundbreaking model adopted by major players like Microsoft and Perplexity AI.

We examine key use cases, fine-tuning methods, and the broader opportunities Deep Seek presents for innovation and automation. Don’t miss discussions on Alibaba’s Qwen 2.5 Max, OpenAI's latest o3 Mini models (32:42) operator implementation (39:37) and more.

Links:
AI wars are basically Chinese in China versus Chinese in the US
https://youtu.be/x9Ekl9Izd38?si=daGbinOqhKzqSDgP 

AI Explained https://youtu.be/FraQpapjQ18?si=8ZY5ImfMKS0qcpJY 

How it works  - from Apple https://www.zdnet.com/article/apple-researchers-reveal-the-secret-sauce-behind-deepseek-ai/

Great examples of DeepSeek 
https://x.com/minchoi/status/1883188761854669147?t=_nbrI2xJveEBJyL7C_jRbA&s=19 

Fine tuning deep seek
https://x.com/_avichawla/status/1884126766132011149?t=d_auYsFvXXEi3t9wo4xIJA&s=19 

Perplexity and deepseek
https://www.searchenginejournal.com/perplexity-ai-deploys-chinese-deepseek-ai-model/538452/ 

Mark's thoughts on the Deep Seek opportunity
https://www.linkedin.com/posts/activity-7290177131089694720-spVj?utm_source=share&utm_medium=member_desktop 

ChatGPT for government 
https://www.theverge.com/news/598852/openai-chatgpt-gov-government-agencies

Microsoft seems to be embracing it though: https://techcrunch.com/2025/01/29/microsoft-brings-a-deepseek-model-to-its-cloud/

Aus Leaders thoughts on Deep Seek
https://www.afr.com/technology/what-australia-s-tech-leaders-think-of-deepseek-s-ai-20250129-p5l7xp

Local Deep Seek : https://www.pcgamer.com/hardware/graphics-cards/today-i-learned-i-can-run-my-very-own-deepseek-r1-chatbot-on-just-usd6-000-of-pc-hardware-and-no-megabucks-nvidia-gpus-required/

Deep Seek avoiding answering certain questions
https://techcrunch.com/2025/01/29/deepseeks-ai-avoids-answering-85-of-prompts-on-sensitive-topics-related-to-china/

Showcasing Qwen2 from Alibaba: https://www.pcgamer.com/hardware/forget-deepseek-r1-apparently-its-now-alibaba-that-has-the-most-powerful-the-cheapest-the-most-everything-est-chatbot/

Transcript

1
01:00:03,725 --> 01:00:05,727
Episode 22, we're back another week.

2
01:00:05,727 --> 01:00:06,894
A lot of stuff that's happened.

3
01:00:07,228 --> 01:00:09,147
It feels like it's been a year already.

4
01:00:10,189 --> 01:00:10,565
What have you done?

5
01:00:10,940 --> 01:00:12,900
Three weeks in a day, uh, three weeks of

6
01:00:12,900 --> 01:00:13,484
like major

7
01:00:13,484 --> 01:00:14,819
announcements, all in one week,

8
01:00:14,819 --> 01:00:16,321
all in one day, it seems like we had

9
01:00:16,321 --> 01:00:18,114
operated, we had deep seek, we had

10
01:00:18,156 --> 01:00:19,949
03 and we were testing it out.

11
01:00:20,199 --> 01:00:22,160
We go through a lot of like AR stuff that

12
01:00:22,160 --> 01:00:23,494
was in the news, but we also

13
01:00:23,661 --> 01:00:25,830
showcase what we have been building.

14
01:00:26,497 --> 01:00:26,914
That's right.

15
01:00:26,914 --> 01:00:27,415
Lots of investments

16
01:00:27,415 --> 01:00:28,416
going in all over the place.

17
01:00:28,583 --> 01:00:29,417
Huge market crashes.

18
01:00:30,126 --> 01:00:31,919
You hear it all first on digital access.

19
01:00:32,295 --> 01:00:32,587
See you soon.

20
01:00:35,590 --> 01:00:37,133
It's like, "Sir, this is a Wendy's."

21
01:00:38,718 --> 01:00:39,552
Is there a Wendy's here?

22
01:00:40,261 --> 01:00:41,679
Ah, yeah. Do we have Wendy's?

23
01:00:41,679 --> 01:00:42,096
Yeah, we don't have

24
01:00:42,096 --> 01:00:43,222
Wendy's, I don't think.

25
01:00:43,848 --> 01:00:44,390
Is that used to?

26
01:00:44,599 --> 01:00:47,268
I think there was like a controversy

27
01:00:47,268 --> 01:00:49,687
around potentially getting a Wendy's.

28
01:00:49,687 --> 01:00:51,230
I swear we used to have Wendy's.

29
01:00:51,606 --> 01:00:52,940
Now we had Hungry Jacks.

30
01:00:53,149 --> 01:00:54,108
We do have Hungry Jacks.

31
01:00:54,108 --> 01:00:55,318
We used to have Hungry Jacks.

32
01:00:55,526 --> 01:00:56,402
I swear we had Wendy's.

33
01:00:56,402 --> 01:00:57,653
Did you get a Wendy's or Hungry Jacks?

34
01:00:57,653 --> 01:00:58,071
Is that what it was?

35
01:00:59,072 --> 01:00:59,906
Wendy's hand...

36
01:00:59,947 --> 01:01:01,532
Look, wendys.com.au.

37
01:01:02,325 --> 01:01:03,659
Yeah. So I'm frustrated with it.

38
01:01:03,659 --> 01:01:03,910
You're kidding.

39
01:01:04,369 --> 01:01:05,536
Oh, that's right. It only

40
01:01:05,536 --> 01:01:06,704
just opened in Bass Hill.

41
01:01:06,996 --> 01:01:07,663
Hello, folks.

42
01:01:07,872 --> 01:01:09,457
Welcome back to episode 22.

43
01:01:10,375 --> 01:01:11,584
I don't know why we're doing this.

44
01:01:12,001 --> 01:01:12,960
The AI told us to.

45
01:01:13,795 --> 01:01:14,337
And more.

46
01:01:15,088 --> 01:01:15,630
We were saying bad

47
01:01:15,630 --> 01:01:16,547
things about Sam Uldman.

48
01:01:17,590 --> 01:01:18,966
And more from like the future.

49
01:01:19,300 --> 01:01:19,884
Point behind the camera.

50
01:01:20,343 --> 01:01:22,220
This is what you can expect from the

51
01:01:22,220 --> 01:01:24,097
future if you don't comply with AI.

52
01:01:24,680 --> 01:01:26,432
Oh, my God. That's the campaign slogan.

53
01:01:27,100 --> 01:01:28,142
There'll be... I'm pretty sure there was

54
01:01:28,142 --> 01:01:30,395
a Black Mirror episode where...

55
01:01:30,728 --> 01:01:32,021
I don't remember it was an AI.

56
01:01:32,063 --> 01:01:34,023
I think it kind of was an AI that was

57
01:01:34,023 --> 01:01:35,566
like running for

58
01:01:35,566 --> 01:01:37,151
parliament or something like that.

59
01:01:37,652 --> 01:01:39,737
And it has a human conduit that...

60
01:01:40,863 --> 01:01:44,450
There's like a little mascot kind of

61
01:01:44,450 --> 01:01:47,328
thing and it ends up like winning.

62
01:01:47,328 --> 01:01:48,121
I've got to take a look.

63
01:01:48,663 --> 01:01:49,205
There were so many

64
01:01:49,205 --> 01:01:50,873
different Black Mirror episodes.

65
01:01:51,749 --> 01:01:53,042
I remember the... I'm

66
01:01:53,042 --> 01:01:54,335
trying to think of the AI ones.

67
01:01:54,335 --> 01:01:54,961
There was ones where there

68
01:01:54,961 --> 01:01:55,920
was like an AI that was...

69
01:01:56,212 --> 01:01:58,381
There was an AI for relationships and...

70
01:01:58,381 --> 01:02:00,675
Yep. That was Hang the DJ.

71
01:02:00,967 --> 01:02:02,510
I can't remember why they named that one.

72
01:02:04,429 --> 01:02:06,013
The AI run for parliament.

73
01:02:06,848 --> 01:02:07,432
Let's look that one up.

74
01:02:07,432 --> 01:02:08,224
Look it up. Look it up.

75
01:02:08,599 --> 01:02:09,392
I'm looking it up.

76
01:02:09,392 --> 01:02:10,601
I'll cut the research.

77
01:02:12,228 --> 01:02:13,271
No, this is all part of it.

78
01:02:13,855 --> 01:02:15,231
Running for parliament.

79
01:02:16,774 --> 01:02:18,151
Black Mirror.

80
01:02:19,235 --> 01:02:19,777
So you're going to

81
01:02:19,777 --> 01:02:20,528
look something up there?

82
01:02:20,528 --> 01:02:21,154
The Waldo moment.

83
01:02:21,571 --> 01:02:23,322
Yeah, an AI candidate in modern politics.

84
01:02:24,323 --> 01:02:27,702
So this explored a character named Waldo

85
01:02:27,702 --> 01:02:28,953
who becomes an unexpected

86
01:02:29,412 --> 01:02:30,872
political face and

87
01:02:30,872 --> 01:02:32,123
force in the real world.

88
01:02:32,373 --> 01:02:32,874
It was AI.

89
01:02:33,249 --> 01:02:35,585
And then in June 2024, there was an AI

90
01:02:35,585 --> 01:02:37,545
candidate, Hank Cancicut,

91
01:02:38,171 --> 01:02:40,631
because I can't speak, an AI candidate

92
01:02:40,631 --> 01:02:41,674
named Steve Endicott.

93
01:02:42,049 --> 01:02:43,301
He was reinventing

94
01:02:43,301 --> 01:02:45,845
democracy, known as AI Steve.

95
01:02:46,053 --> 01:02:47,388
The bot uses Endicott's

96
01:02:47,388 --> 01:02:48,514
actual voice, but it was a bot.

97
01:02:49,265 --> 01:02:50,224
Yeah, AI Steve.

98
01:02:51,017 --> 01:02:52,351
That is crazy.

99
01:02:53,144 --> 01:02:54,228
Scuba Steve's out there.

100
01:02:54,645 --> 01:02:55,563
Of course it's an independent.

101
01:02:55,563 --> 01:02:56,105
It's a real story.

102
01:02:56,439 --> 01:02:58,232
Yeah. So June 2024, last year.

103
01:02:58,399 --> 01:03:00,276
With an AI running for as a candidate.

104
01:03:00,276 --> 01:03:02,111
Yeah. So based on a real guy and his real

105
01:03:02,111 --> 01:03:03,946
voice cloned for the AI.

106
01:03:04,030 --> 01:03:04,363
What?

107
01:03:04,614 --> 01:03:04,822
Yeah.

108
01:03:05,072 --> 01:03:06,407
So Black Mirror is coming to life.

109
01:03:06,616 --> 01:03:08,993
Yeah. I mean, who knows how well he did?

110
01:03:08,993 --> 01:03:09,744
Let's have a look.

111
01:03:10,578 --> 01:03:12,413
Steve Endicott results.

112
01:03:13,498 --> 01:03:14,665
You might not have done that well.

113
01:03:15,666 --> 01:03:17,752
Virtual candidate finishes last in polls.

114
01:03:18,252 --> 01:03:19,754
But hey, you got to start from somewhere.

115
01:03:19,754 --> 01:03:20,671
He was on there, right?

116
01:03:22,590 --> 01:03:23,925
That is, that's fantastic.

117
01:03:24,217 --> 01:03:25,927
He only got 179 votes.

118
01:03:26,177 --> 01:03:27,678
I mean, come on.

119
01:03:28,137 --> 01:03:28,846
How's your week been?

120
01:03:29,847 --> 01:03:30,932
Pretty good, man.

121
01:03:30,932 --> 01:03:31,807
Pretty good. We've

122
01:03:31,807 --> 01:03:32,892
done a whole lot of stuff.

123
01:03:33,267 --> 01:03:34,477
We came back.

124
01:03:34,477 --> 01:03:38,564
We did the first in, I think, over a

125
01:03:38,564 --> 01:03:40,650
year, maybe even close to two years.

126
01:03:41,192 --> 01:03:44,320
Twitter spaces or X spaces.

127
01:03:45,905 --> 01:03:47,823
So, yeah, we had a conversation online.

128
01:03:48,032 --> 01:03:49,534
You can hear some of it here.

129
01:03:49,784 --> 01:03:50,993
So I don't know if this will play.

130
01:03:51,661 --> 01:03:52,703
It's much better on the phone.

131
01:03:53,246 --> 01:03:54,497
But we used to do these

132
01:03:54,497 --> 01:03:56,624
every Monday after work.

133
01:03:56,791 --> 01:03:59,961
And so, yeah, I can't fast forward it.

134
01:04:01,003 --> 01:04:01,462
Here we go.

135
01:04:02,755 --> 01:04:02,964
Yeah.

136
01:04:04,423 --> 01:04:04,674
Yeah.

137
01:04:05,341 --> 01:04:05,925
You have different

138
01:04:05,925 --> 01:04:08,636
people speaking and joining.

139
01:04:09,011 --> 01:04:10,930
And we had just folks in the community.

140
01:04:11,430 --> 01:04:12,974
And it's part of our plan to actually

141
01:04:12,974 --> 01:04:13,891
bring back a lot more

142
01:04:13,891 --> 01:04:15,101
digital online stuff

143
01:04:15,101 --> 01:04:17,436
to SD5 because put it this way, there's

144
01:04:17,436 --> 01:04:18,729
just far more reach.

145
01:04:18,729 --> 01:04:20,273
When I look at the numbers, yeah, you get

146
01:04:20,273 --> 01:04:22,191
like 50 to 100 people at an event.

147
01:04:22,400 --> 01:04:25,528
You run 14, 20, whatever it is of those.

148
01:04:25,861 --> 01:04:27,530
That's only a few thousand people.

149
01:04:28,197 --> 01:04:29,282
Far more reach and far

150
01:04:29,282 --> 01:04:31,158
more regular doing online.

151
01:04:31,158 --> 01:04:32,994
We'll still do some of the IRL events.

152
01:04:32,994 --> 01:04:33,786
That's going to happen.

153
01:04:33,786 --> 01:04:34,453
But I think we're just going

154
01:04:34,453 --> 01:04:35,871
to do a lot more online stuff.

155
01:04:36,163 --> 01:04:36,539
Yeah, great.

156
01:04:36,831 --> 01:04:39,166
Because people are really jumping on it.

157
01:04:39,166 --> 01:04:40,751
And I was surprised that even get I think

158
01:04:40,751 --> 01:04:41,627
we had more inclusion.

159
01:04:42,128 --> 01:04:42,420
It's good.

160
01:04:42,712 --> 01:04:43,296
More inclusion.

161
01:04:43,296 --> 01:04:44,130
So we did that.

162
01:04:44,338 --> 01:04:45,923
And then another big thing, Chris, is

163
01:04:45,923 --> 01:04:48,050
that we've got our online kickoff,

164
01:04:49,635 --> 01:04:50,595
which is on Tuesday.

165
01:04:50,928 --> 01:04:51,470
So I don't know when

166
01:04:51,470 --> 01:04:52,513
this video is coming out.

167
01:04:52,513 --> 01:04:54,181
It might be on Tuesday that you're seeing

168
01:04:54,181 --> 01:04:54,890
this video in the morning.

169
01:04:55,057 --> 01:04:56,851
But we've got our national online kickoff

170
01:04:56,851 --> 01:04:58,311
where we've got

171
01:04:58,311 --> 01:04:59,812
national and international

172
01:04:59,979 --> 01:05:00,771
guests and there's a bit

173
01:05:00,771 --> 01:05:02,064
of flavor of blockchain.

174
01:05:02,773 --> 01:05:04,400
It's for the OzD5 Association, but

175
01:05:04,400 --> 01:05:04,942
there's going to be

176
01:05:04,942 --> 01:05:06,402
some AI elements in there.

177
01:05:06,402 --> 01:05:08,863
There's a project from overseas that is

178
01:05:08,863 --> 01:05:10,531
using blockchain to

179
01:05:10,531 --> 01:05:11,991
help secure the data or

180
01:05:11,991 --> 01:05:12,783
bring trust to the

181
01:05:12,783 --> 01:05:14,201
data that's used in AI.

182
01:05:15,161 --> 01:05:16,037
They're former NIR

183
01:05:16,037 --> 01:05:18,956
founders and former NIR engineers.

184
01:05:20,082 --> 01:05:22,126
NIR is being one of the key blockchains

185
01:05:22,126 --> 01:05:23,669
doing stuff in the AI space.

186
01:05:23,669 --> 01:05:24,503
And so they've left there.

187
01:05:24,545 --> 01:05:26,172
They've formed this protocol.

188
01:05:26,422 --> 01:05:28,090
And so they'll be talking.

189
01:05:28,341 --> 01:05:29,050
Arturo will be there.

190
01:05:30,217 --> 01:05:31,385
And yeah, some great stuff.

191
01:05:32,053 --> 01:05:32,803
Arturo is going to be there.

192
01:05:32,970 --> 01:05:36,140
Leah, who was formerly OzD5 in Sydney and

193
01:05:36,140 --> 01:05:37,683
stuff, and she's over in Dubai.

194
01:05:37,683 --> 01:05:39,894
So we're bringing back the band together

195
01:05:39,894 --> 01:05:40,478
and learning about

196
01:05:40,478 --> 01:05:41,812
some others in the space.

197
01:05:42,063 --> 01:05:43,898
Peter Jing from Transhuman, Lucas Cullen,

198
01:05:44,982 --> 01:05:46,067
Ben Ritchie, et cetera.

199
01:05:46,067 --> 01:05:47,443
So yeah, that's going to be really good.

200
01:05:47,443 --> 01:05:50,363
But I've had a whole lot of stuff on in

201
01:05:50,363 --> 01:05:51,447
terms of just planning that.

202
01:05:51,697 --> 01:05:52,948
But yeah, that's been my week.

203
01:05:52,948 --> 01:05:53,324
What about you?

204
01:05:54,033 --> 01:05:55,534
My week's been interesting because we're

205
01:05:55,534 --> 01:05:57,370
obviously in the process of thinking

206
01:05:57,370 --> 01:05:58,829
about how we build this new AI tool

207
01:05:59,747 --> 01:06:01,332
beyond its proof of concept.

208
01:06:01,540 --> 01:06:02,833
So I've just been meeting with people,

209
01:06:02,833 --> 01:06:05,044
having discussions, understanding how we

210
01:06:05,044 --> 01:06:05,753
can get some funding.

211
01:06:06,170 --> 01:06:07,713
Did you meet some AI people as well?

212
01:06:08,005 --> 01:06:09,256
No, I haven't met any AI people.

213
01:06:09,256 --> 01:06:10,758
I've been talking with AI a lot, like,

214
01:06:10,758 --> 01:06:11,717
what do they do for this?

215
01:06:11,717 --> 01:06:12,760
And how should I think about this?

216
01:06:12,760 --> 01:06:14,136
So, I mean, yes, technically, actually.

217
01:06:14,345 --> 01:06:15,721
Yeah, this should be like AI developer

218
01:06:15,721 --> 01:06:18,599
that you chat with and they're the ones

219
01:06:18,599 --> 01:06:19,642
that build it or something.

220
01:06:20,142 --> 01:06:21,185
Why not?

221
01:06:21,394 --> 01:06:21,519
Right.

222
01:06:21,519 --> 01:06:23,270
I mean, look, there's Devon.

223
01:06:24,063 --> 01:06:25,606
Which is the thing where I think you pay

224
01:06:25,606 --> 01:06:27,024
a thousand dollars a month or something

225
01:06:27,024 --> 01:06:28,484
like that, probably a thousand US.

226
01:06:29,276 --> 01:06:31,112
And when I think about it, it's like that

227
01:06:31,112 --> 01:06:32,863
pricing is not actually that bad.

228
01:06:33,155 --> 01:06:34,365
And I'm going to tell you why, because

229
01:06:34,365 --> 01:06:36,242
like using tools like Bolt, like it does

230
01:06:36,242 --> 01:06:37,576
get up there in terms of pricing.

231
01:06:37,576 --> 01:06:40,246
When it's for work, it's worth, you know,

232
01:06:40,663 --> 01:06:42,373
playing the extra, paying for

233
01:06:42,373 --> 01:06:43,666
the highest amount and stuff.

234
01:06:43,833 --> 01:06:46,335
And you're the one driving it and just

235
01:06:46,335 --> 01:06:48,212
seeing how quickly cost can rack up,

236
01:06:48,212 --> 01:06:49,797
especially as you're

237
01:06:49,797 --> 01:06:50,673
getting it to do stuff.

238
01:06:51,132 --> 01:06:51,632
And sure, there's

239
01:06:51,632 --> 01:06:52,675
probably a more efficient way.

240
01:06:52,717 --> 01:06:56,178
But I can see value in, you know, using

241
01:06:56,178 --> 01:06:58,139
tools like that that seem expensive, like

242
01:06:58,139 --> 01:06:59,974
Pro Chat GPT, Pro,

243
01:06:59,974 --> 01:07:01,350
Plus or whatever it is.

244
01:07:01,350 --> 01:07:01,767
Plus or Pro?

245
01:07:01,767 --> 01:07:02,435
Plus or Pro?

246
01:07:02,435 --> 01:07:03,144
What were you saying before?

247
01:07:03,352 --> 01:07:04,937
This should be like a free...

248
01:07:05,187 --> 01:07:05,312
View plans.

249
01:07:05,312 --> 01:07:05,563
Here we go.

250
01:07:05,730 --> 01:07:06,063
Pro.

251
01:07:06,063 --> 01:07:06,647
No, it is it.

252
01:07:06,647 --> 01:07:08,315
So plus is the 20, pro is the 200.

253
01:07:08,691 --> 01:07:11,610
I would have had free a little bit and

254
01:07:11,610 --> 01:07:12,820
then a whole lot more.

255
01:07:13,070 --> 01:07:13,446
And then you got...

256
01:07:13,446 --> 01:07:14,113
That's the three tiers.

257
01:07:14,113 --> 01:07:15,281
You got business plans for teams.

258
01:07:15,531 --> 01:07:16,115
So what was it again?

259
01:07:16,574 --> 01:07:17,408
Pro is the highest.

260
01:07:17,742 --> 01:07:18,409
Pro is the highest.

261
01:07:18,409 --> 01:07:18,492
You were...

262
01:07:18,492 --> 01:07:18,743
Trust.

263
01:07:19,285 --> 01:07:19,493
Trust.

264
01:07:19,952 --> 01:07:20,161
Yeah.

265
01:07:20,369 --> 01:07:20,578
Fresh.

266
01:07:20,619 --> 01:07:21,162
Freshy.

267
01:07:21,162 --> 01:07:21,537
As in you are.

268
01:07:21,829 --> 01:07:22,538
Pro is your brush.

269
01:07:23,038 --> 01:07:23,247
Yeah.

270
01:07:23,330 --> 01:07:25,541
So how's it going in terms of like, you

271
01:07:25,541 --> 01:07:26,876
know, having these conversations?

272
01:07:27,126 --> 01:07:27,626
Yeah, interesting.

273
01:07:27,626 --> 01:07:29,003
It's always fun having these chats and

274
01:07:29,003 --> 01:07:30,921
like seeing who's excited to do things.

275
01:07:31,130 --> 01:07:33,674
Like it's interesting even, I mean, in

276
01:07:33,674 --> 01:07:36,844
the market, all the real AI devs are

277
01:07:36,844 --> 01:07:38,262
getting eaten up really quickly.

278
01:07:38,471 --> 01:07:40,848
So finding someone who's working for

279
01:07:40,848 --> 01:07:42,224
themselves, wants to, you know, get

280
01:07:42,224 --> 01:07:42,975
involved in a startup

281
01:07:42,975 --> 01:07:44,894
is, it's not an easy task.

282
01:07:44,894 --> 01:07:45,144
No.

283
01:07:45,144 --> 01:07:45,478
Especially someone with

284
01:07:45,478 --> 01:07:46,562
some experience, I understand.

285
01:07:46,562 --> 01:07:47,563
Because it's such a new thing.

286
01:07:47,563 --> 01:07:48,189
There's not a lot of

287
01:07:48,189 --> 01:07:49,231
developers out there.

288
01:07:49,732 --> 01:07:50,691
There are diamond does it at the moment.

289
01:07:51,025 --> 01:07:52,568
Well, hopefully not a

290
01:07:52,568 --> 01:07:53,527
diamond does it and stuff.

291
01:07:53,527 --> 01:07:54,820
There's still a good price on them.

292
01:07:54,820 --> 01:07:56,405
But I mean, the prices are going down.

293
01:07:56,697 --> 01:07:56,906
So...

294
01:07:56,906 --> 01:07:58,324
The availability of them.

295
01:07:58,324 --> 01:07:58,407
Yes.

296
01:07:58,407 --> 01:07:59,492
Not the cost of them.

297
01:07:59,492 --> 01:07:59,784
Yes.

298
01:07:59,784 --> 01:08:00,326
No, the cost.

299
01:08:00,326 --> 01:08:00,993
That's for Chris, man.

300
01:08:00,993 --> 01:08:02,536
That's what he expects to pay you.

301
01:08:02,536 --> 01:08:03,162
So if you want to help

302
01:08:03,162 --> 01:08:05,831
Chris for a dime, but 12 of you.

303
01:08:06,207 --> 01:08:07,208
12 people for one.

304
01:08:08,042 --> 01:08:09,627
But really interesting times as well.

305
01:08:09,877 --> 01:08:10,878
Well, I've been enjoying even just

306
01:08:10,878 --> 01:08:13,756
looking at the news, just around what's

307
01:08:13,756 --> 01:08:14,423
happening in the world.

308
01:08:14,632 --> 01:08:15,174
I mean, I was just

309
01:08:15,174 --> 01:08:17,468
fucking just incredibly chaotic.

310
01:08:17,468 --> 01:08:18,010
It's insane.

311
01:08:18,010 --> 01:08:18,344
It's great.

312
01:08:18,511 --> 01:08:20,888
Trump is feeling my

313
01:08:20,888 --> 01:08:22,306
week with happy smiles.

314
01:08:23,057 --> 01:08:23,933
Or like, what?

315
01:08:24,099 --> 01:08:24,433
Of disaster.

316
01:08:24,767 --> 01:08:25,392
But it's hilarious.

317
01:08:25,768 --> 01:08:27,770
It is hilarious just what is going on.

318
01:08:27,770 --> 01:08:29,271
Welcome back to drama.

319
01:08:30,523 --> 01:08:30,731
2025.

320
01:08:31,232 --> 01:08:33,025
Like three planes have collided, like

321
01:08:33,025 --> 01:08:34,068
crashed in the last two

322
01:08:34,068 --> 01:08:35,027
weeks or something silly.

323
01:08:35,402 --> 01:08:35,653
Crazy.

324
01:08:35,986 --> 01:08:38,113
RFP, you know, really sad

325
01:08:38,113 --> 01:08:39,782
stuff, but like jumping on.

326
01:08:40,032 --> 01:08:41,033
Joe, who's to blame with it all?

327
01:08:41,408 --> 01:08:42,409
Oh, yeah, it's DEI.

328
01:08:42,409 --> 01:08:42,785
That's...

329
01:08:42,785 --> 01:08:42,952
Yeah.

330
01:08:43,285 --> 01:08:43,536
According to...

331
01:08:43,786 --> 01:08:44,453
World according to Trump.

332
01:08:44,954 --> 01:08:45,162
Yeah.

333
01:08:45,371 --> 01:08:45,496
According to Trump.

334
01:08:45,538 --> 01:08:48,207
If it wasn't for data engineering and

335
01:08:48,207 --> 01:08:50,417
ingenuity, I mean, that those are

336
01:08:50,417 --> 01:08:51,961
definitely not things that we want in

337
01:08:51,961 --> 01:08:52,628
this administration.

338
01:08:52,920 --> 01:08:54,130
No, not at all.

339
01:08:54,380 --> 01:08:54,839
So it was...

340
01:08:54,839 --> 01:08:55,548
It is crazy.

341
01:08:55,548 --> 01:08:56,674
One of his recent comments...

342
01:08:56,674 --> 01:08:57,383
Chumsick conclusions.

343
01:08:57,675 --> 01:08:59,552
Yeah, one of his recent comments, I can't

344
01:08:59,552 --> 01:09:00,761
remember which of the airline it is that

345
01:09:00,761 --> 01:09:02,096
was responsible for it all, but he's like

346
01:09:02,096 --> 01:09:02,972
saying how there was...

347
01:09:03,597 --> 01:09:05,558
It's that they are allowing people with

348
01:09:05,558 --> 01:09:06,809
mental disabilities in

349
01:09:06,809 --> 01:09:07,810
to control this stuff.

350
01:09:07,810 --> 01:09:08,394
And that's why these

351
01:09:08,394 --> 01:09:09,103
planes are going down.

352
01:09:09,103 --> 01:09:09,562
It's just like...

353
01:09:10,479 --> 01:09:11,355
That's just not...

354
01:09:11,355 --> 01:09:11,939
This is just not...

355
01:09:11,981 --> 01:09:13,190
The big problem is that

356
01:09:13,190 --> 01:09:14,483
y'all are enabling it.

357
01:09:14,692 --> 01:09:15,776
Like the...

358
01:09:16,652 --> 01:09:18,070
I'm sure that there's a lot of folks that

359
01:09:18,070 --> 01:09:20,739
voted not for the intellect, but just

360
01:09:20,739 --> 01:09:23,576
because of the outcomes of whatever is

361
01:09:23,576 --> 01:09:25,244
going to be, they think, changing.

362
01:09:25,578 --> 01:09:26,537
But the problem is

363
01:09:26,537 --> 01:09:27,663
that it's being enabled.

364
01:09:28,122 --> 01:09:30,291
You know, people believe this stuff that

365
01:09:30,291 --> 01:09:31,208
it's like true because

366
01:09:31,208 --> 01:09:32,084
they want to believe it.

367
01:09:32,459 --> 01:09:32,835
Well, he's supposed to be someone...

368
01:09:33,043 --> 01:09:34,503
He's a person that's

369
01:09:34,503 --> 01:09:35,504
supposed to be believed in.

370
01:09:35,504 --> 01:09:36,463
So of course they're going to...

371
01:09:36,463 --> 01:09:37,131
Yeah, it's crazy.

372
01:09:37,256 --> 01:09:37,923
You know, it's interesting.

373
01:09:37,923 --> 01:09:39,133
Just just on the topic, I don't have an

374
01:09:39,133 --> 01:09:40,301
article for this, but...

375
01:09:41,135 --> 01:09:43,095
There was proof that if you've got like

376
01:09:43,095 --> 01:09:45,431
these extremist kind of views, let's say,

377
01:09:45,848 --> 01:09:47,975
you know, whatever side of the spectrum

378
01:09:47,975 --> 01:09:51,270
it's on, using AI actually helps.

379
01:09:51,270 --> 01:09:54,273
As in it helps you break down some of

380
01:09:54,273 --> 01:09:58,277
these like wrongly perceive things.

381
01:09:58,611 --> 01:09:59,653
Like, and no, no, no, no.

382
01:09:59,653 --> 01:10:02,114
Like there's research that has this and

383
01:10:02,114 --> 01:10:03,365
Ethan Molek, I'm pretty sure is the one

384
01:10:03,365 --> 01:10:04,992
that's put it out there, where it can

385
01:10:04,992 --> 01:10:06,577
help bring back balance and

386
01:10:06,577 --> 01:10:08,078
help you see the other side.

387
01:10:08,329 --> 01:10:10,080
And I don't know how, sorry, I don't know

388
01:10:10,080 --> 01:10:11,707
how they've actually done the test.

389
01:10:12,291 --> 01:10:14,043
You know, what kind of closed environment

390
01:10:14,043 --> 01:10:16,378
or like how they, you know, surveys.

391
01:10:16,378 --> 01:10:18,005
It's like there's so many biases that you

392
01:10:18,005 --> 01:10:20,507
can do in surveys, but potentially I

393
01:10:20,507 --> 01:10:22,843
could see AI actually bringing more

394
01:10:22,843 --> 01:10:24,178
balance to someone's like thoughts.

395
01:10:24,178 --> 01:10:28,223
The caution there, and it's the same

396
01:10:28,223 --> 01:10:30,434
caution that you where people get into

397
01:10:30,434 --> 01:10:33,646
these spirals around this particular dark

398
01:10:33,646 --> 01:10:35,314
theory, that particular dark theory,

399
01:10:35,898 --> 01:10:38,400
is they tend to search by the negative.

400
01:10:38,859 --> 01:10:39,068
Yeah.

401
01:10:39,276 --> 01:10:41,820
And I think if you ask the AI about the

402
01:10:41,820 --> 01:10:44,073
negative, so for example, let's just do

403
01:10:44,073 --> 01:10:45,407
the funny one, this is the flat earth.

404
01:10:45,991 --> 01:10:46,200
Yeah.

405
01:10:46,533 --> 01:10:48,869
You know, if you go, why is the earth

406
01:10:48,869 --> 01:10:50,913
flat rather than is the earth

407
01:10:50,913 --> 01:10:53,082
or why, how is the earth round?

408
01:10:53,374 --> 01:10:53,582
Yeah.

409
01:10:53,958 --> 01:10:55,668
It's the same thing if you two go to AI,

410
01:10:55,876 --> 01:10:59,171
go why is the earth, it will hallucinate

411
01:10:59,171 --> 01:10:59,880
the negative because

412
01:10:59,880 --> 01:11:01,382
that's how you've queried it.

413
01:11:01,382 --> 01:11:02,925
Of course, I think, and this is the

414
01:11:02,925 --> 01:11:03,842
thing, like, as I said...

415
01:11:03,884 --> 01:11:04,885
So it could feed the fuel

416
01:11:04,885 --> 01:11:06,512
as even more than as well.

417
01:11:06,512 --> 01:11:08,097
It could feed the fuel.

418
01:11:08,097 --> 01:11:08,889
That's exactly it.

419
01:11:09,390 --> 01:11:09,682
Fuel the...

420
01:11:10,140 --> 01:11:11,016
You know what I mean.

421
01:11:11,850 --> 01:11:15,604
Chris has backwards bias and it is one of

422
01:11:15,604 --> 01:11:16,897
those things where he...

423
01:11:16,897 --> 01:11:17,523
Rick Jels had my coffee.

424
01:11:17,606 --> 01:11:18,649
...wrong thing says.

425
01:11:19,692 --> 01:11:22,403
So he has a coffee...

426
01:11:22,403 --> 01:11:23,028
Fueled the fire.

427
01:11:23,028 --> 01:11:23,654
...had his.

428
01:11:23,654 --> 01:11:24,405
Fire the fuel.

429
01:11:24,405 --> 01:11:25,906
So fire the fuel.

430
01:11:26,240 --> 01:11:27,157
You should fire the fuel.

431
01:11:27,908 --> 01:11:30,202
But I think I'm going to guess, I am

432
01:11:30,202 --> 01:11:30,953
going to guess that what

433
01:11:30,953 --> 01:11:32,037
they've done is that...

434
01:11:32,705 --> 01:11:33,497
Because you can do this

435
01:11:33,497 --> 01:11:34,999
with custom prompts, right?

436
01:11:34,999 --> 01:11:35,874
That's it, kind of seems.

437
01:11:35,874 --> 01:11:38,544
Whatever you ask, get it to actually give

438
01:11:38,544 --> 01:11:40,129
you the opposing kind of views.

439
01:11:40,337 --> 01:11:41,588
And if that's what they've done, then

440
01:11:41,588 --> 01:11:42,798
yeah, I can see how they've done it.

441
01:11:42,798 --> 01:11:44,133
And obviously you're right.

442
01:11:44,133 --> 01:11:45,926
Like, we have to be very careful with

443
01:11:45,926 --> 01:11:47,970
that because we saw early on people

444
01:11:47,970 --> 01:11:50,556
creating and going and spiraling down,

445
01:11:50,556 --> 01:11:51,974
writing their own books and doing all

446
01:11:51,974 --> 01:11:53,892
these other crazy things where it just

447
01:11:53,892 --> 01:11:57,021
fed the fire with fuel.

448
01:11:59,148 --> 01:12:00,149
I'm being even careful.

449
01:12:00,315 --> 01:12:00,941
Just you, man.

450
01:12:00,941 --> 01:12:01,608
Just embed the fuel.

451
01:12:02,109 --> 01:12:02,359
We should...

452
01:12:02,943 --> 01:12:05,612
I love how they can say something that

453
01:12:05,612 --> 01:12:07,573
sounds so intelligent in a moment and

454
01:12:07,573 --> 01:12:09,366
then end it with absolute stupidity and

455
01:12:09,366 --> 01:12:11,201
then just like completely, just

456
01:12:11,201 --> 01:12:12,703
completely ruin any sort of

457
01:12:12,703 --> 01:12:13,662
smile that I would have said.

458
01:12:13,662 --> 01:12:14,538
It's like, "Oh, it sounds smart."

459
01:12:14,538 --> 01:12:15,414
Oh, no, actually it's a dumb idea.

460
01:12:15,414 --> 01:12:17,291
AI will turn it into

461
01:12:17,291 --> 01:12:18,250
something even better.

462
01:12:19,209 --> 01:12:20,294
Like, it will find...

463
01:12:20,294 --> 01:12:21,670
This is the greatest thing, like AI as a

464
01:12:21,670 --> 01:12:22,629
translation tool, but

465
01:12:22,629 --> 01:12:24,715
also a meaning finding tool.

466
01:12:24,715 --> 01:12:26,508
Remember when like Notebook LM was out

467
01:12:26,508 --> 01:12:27,134
and we were doing the

468
01:12:27,134 --> 01:12:28,052
podcast stuff and...

469
01:12:28,844 --> 01:12:29,887
I see people talking

470
01:12:29,887 --> 01:12:30,929
about that again online.

471
01:12:31,180 --> 01:12:32,264
We have an excellent episode up ahead here.

472
01:12:32,264 --> 01:12:32,473
It's good.

473
01:12:33,098 --> 01:12:35,726
But it was great at giving you insights

474
01:12:35,726 --> 01:12:37,811
into like a really mundane topic and

475
01:12:37,811 --> 01:12:38,645
turning into a podcast.

476
01:12:38,896 --> 01:12:39,438
I was insane.

477
01:12:39,772 --> 01:12:42,524
I remember how I fed it nothing but the

478
01:12:42,524 --> 01:12:45,861
phrase McDonald's McDonald's, Kentucky

479
01:12:45,861 --> 01:12:47,613
Fried Chicken and a Pizza Hut.

480
01:12:47,613 --> 01:12:49,823
And I just had pages of that and I fed it

481
01:12:49,823 --> 01:12:51,867
in and it was like, there must be some

482
01:12:51,867 --> 01:12:53,118
deep meaning behind this.

483
01:12:53,118 --> 01:12:53,786
Is it a comment?

484
01:12:54,328 --> 01:12:57,206
Is on society and commercialization?

485
01:12:57,831 --> 01:13:00,209
And the consumerization of society and it

486
01:13:00,209 --> 01:13:00,959
was just finding meaning

487
01:13:00,959 --> 01:13:03,003
that was never intended there.

488
01:13:03,003 --> 01:13:04,254
But that's the great thing about AI.

489
01:13:04,254 --> 01:13:05,589
So you could say something like what you

490
01:13:05,589 --> 01:13:08,133
said and we will get the AI to turn it

491
01:13:08,133 --> 01:13:08,884
into something even better.

492
01:13:08,884 --> 01:13:09,551
It will be deep.

493
01:13:09,843 --> 01:13:10,552
And that'll be there.

494
01:13:10,761 --> 01:13:11,845
Again, that will fuel the feed.

495
01:13:13,806 --> 01:13:14,264
We're going to make

496
01:13:14,264 --> 01:13:15,724
t-shirts fuel the feed.

497
01:13:16,433 --> 01:13:17,267
You saw it here first.

498
01:13:17,267 --> 01:13:17,851
You just filled whatever.

499
01:13:18,185 --> 01:13:19,812
2026, Chris is running

500
01:13:19,812 --> 01:13:20,979
for local parliament.

501
01:13:21,563 --> 01:13:22,439
Well, with staying super

502
01:13:22,439 --> 01:13:23,315
shit like that, I might win.

503
01:13:23,690 --> 01:13:24,274
Dude, we should.

504
01:13:24,608 --> 01:13:24,983
Yeah, OK.

505
01:13:25,442 --> 01:13:26,068
Let's see.

506
01:13:26,068 --> 01:13:27,402
We might even create a

507
01:13:27,402 --> 01:13:28,570
Democratic candidate.

508
01:13:30,239 --> 01:13:31,490
Yeah, I think I think it'd be great.

509
01:13:31,949 --> 01:13:32,116
Done.

510
01:13:33,450 --> 01:13:34,284
Get a feed, Chris.

511
01:13:34,409 --> 01:13:35,035
What else you got?

512
01:13:35,202 --> 01:13:38,247
What's continuing to fuel the feed?

513
01:13:38,580 --> 01:13:38,789
Yeah.

514
01:13:39,498 --> 01:13:40,833
What I mean, we can't not

515
01:13:40,833 --> 01:13:42,084
talk about the biggest thing.

516
01:13:42,501 --> 01:13:43,210
Operator or deep seek?

517
01:13:43,669 --> 01:13:46,171
Well, I mean, operator is it's fine.

518
01:13:46,922 --> 01:13:47,548
It's out there.

519
01:13:48,006 --> 01:13:48,674
Now it's pretty big.

520
01:13:48,674 --> 01:13:49,383
Not not many people.

521
01:13:49,383 --> 01:13:50,592
I want to I want to do the first the

522
01:13:50,592 --> 01:13:51,343
first big thing because

523
01:13:51,343 --> 01:13:52,678
operator came second in that.

524
01:13:53,053 --> 01:13:53,262
OK.

525
01:13:53,345 --> 01:13:53,762
Deep seek.

526
01:13:53,804 --> 01:13:54,763
Deep seek.

527
01:13:54,972 --> 01:13:56,598
Like, man, let's get the let's get the

528
01:13:56,598 --> 01:13:57,599
big one out of the way.

529
01:13:57,850 --> 01:14:01,770
Yeah, the one miles ride for everyone who

530
01:14:01,770 --> 01:14:03,105
is listening in deep seek.

531
01:14:03,105 --> 01:14:04,731
If you don't know about it, V3 they

532
01:14:04,731 --> 01:14:05,941
launched back last year.

533
01:14:05,941 --> 01:14:08,735
I think it was November, December was

534
01:14:08,735 --> 01:14:10,988
when was it Chinese AI company.

535
01:14:11,280 --> 01:14:12,406
It was pretty good at the time.

536
01:14:12,406 --> 01:14:12,823
Open source.

537
01:14:13,323 --> 01:14:15,826
And then over the last couple of weeks,

538
01:14:15,826 --> 01:14:16,618
two or three weeks ago,

539
01:14:16,618 --> 01:14:18,829
they've launched their

540
01:14:18,829 --> 01:14:21,498
one version of the

541
01:14:21,498 --> 01:14:22,082
model, which was an update.

542
01:14:22,124 --> 01:14:24,293
It is a reasoning, a new reasoning model.

543
01:14:25,085 --> 01:14:27,171
And it has pretty much just shocked the

544
01:14:27,171 --> 01:14:30,090
entire industry from everything from

545
01:14:30,090 --> 01:14:32,092
Nvidia stocks down to open AIs,

546
01:14:32,759 --> 01:14:34,595
evaluations, putting into question.

547
01:14:34,803 --> 01:14:35,012
Yeah.

548
01:14:35,721 --> 01:14:38,640
What are you know, what what investment

549
01:14:38,640 --> 01:14:39,516
actually needs to be made

550
01:14:39,516 --> 01:14:41,476
in the market to to fuel

551
01:14:41,476 --> 01:14:43,437
the to fuel the feed of these tools?

552
01:14:43,854 --> 01:14:44,855
How do you fuel the

553
01:14:44,855 --> 01:14:48,400
feed of your AI portfolio?

554
01:14:49,026 --> 01:14:49,735
You know, it's really interesting.

555
01:14:49,776 --> 01:14:51,612
I think the market over corrected and

556
01:14:51,612 --> 01:14:54,531
coming from a finance

557
01:14:54,531 --> 01:14:56,241
background, there's reasons for

558
01:14:56,241 --> 01:14:57,701
that apart from like the headline that

559
01:14:57,701 --> 01:14:59,786
Nvidia was like down 16, 17

560
01:14:59,786 --> 01:15:01,246
percent and being one of the

561
01:15:01,246 --> 01:15:02,122
biggest companies.

562
01:15:02,789 --> 01:15:04,291
I think at one point it was the biggest

563
01:15:04,291 --> 01:15:05,125
and it's always like up

564
01:15:05,125 --> 01:15:06,251
around there because of all the

565
01:15:06,251 --> 01:15:08,503
AI fuel that's been feeding it.

566
01:15:09,046 --> 01:15:10,339
I'm just going to keep going with it.

567
01:15:11,131 --> 01:15:14,343
But it like a billion billion billion

568
01:15:14,343 --> 01:15:15,886
dollar company losing 17

569
01:15:15,886 --> 01:15:16,970
percent is a huge market

570
01:15:16,970 --> 01:15:17,262
loss.

571
01:15:17,262 --> 01:15:17,888
This is the biggest.

572
01:15:18,096 --> 01:15:20,182
They gain nine percent as it corrected.

573
01:15:20,182 --> 01:15:20,891
It bounced back.

574
01:15:20,891 --> 01:15:22,601
But it's not quite where it was before

575
01:15:22,601 --> 01:15:23,810
because people are still confused.

576
01:15:24,186 --> 01:15:26,021
And, you know, the markets trade.

577
01:15:26,230 --> 01:15:27,397
A lot of it is robotic.

578
01:15:27,773 --> 01:15:29,691
And what I mean is that there's a lot of

579
01:15:29,691 --> 01:15:32,152
AI, you know, normal

580
01:15:32,152 --> 01:15:33,570
ordinary machine learning type

581
01:15:33,612 --> 01:15:35,322
AI and maybe generative AI that is

582
01:15:35,322 --> 01:15:36,365
running the trading

583
01:15:36,365 --> 01:15:38,200
algorithms for a lot of these places.

584
01:15:38,200 --> 01:15:39,993
And yes, there's humans and they would

585
01:15:39,993 --> 01:15:40,661
have been looking for

586
01:15:40,661 --> 01:15:41,828
things that are like negative

587
01:15:41,870 --> 01:15:43,372
sentiment and stuff in the news and

588
01:15:43,372 --> 01:15:44,873
trading off the back of that.

589
01:15:45,332 --> 01:15:49,711
Now, if if they've done that and it's

590
01:15:49,711 --> 01:15:50,754
very likely that they have

591
01:15:50,754 --> 01:15:53,382
the reason why things move so

592
01:15:53,382 --> 01:15:54,508
fast, because as soon as there's new

593
01:15:54,508 --> 01:15:56,760
information in, they will trade on it

594
01:15:56,760 --> 01:15:59,346
right faster and at much

595
01:15:59,346 --> 01:16:00,347
more scale than humans.

596
01:16:00,722 --> 01:16:03,850
So part of it is, you know, the negative

597
01:16:03,850 --> 01:16:05,477
sentiment that's out there and potential

598
01:16:05,477 --> 01:16:06,812
misinterpretations of that.

599
01:16:07,145 --> 01:16:10,065
So and then people also seeing things go

600
01:16:10,065 --> 01:16:11,233
down and then the human

601
01:16:11,233 --> 01:16:12,651
collective just coming in and, you

602
01:16:12,651 --> 01:16:13,610
know, jumping onto that.

603
01:16:13,610 --> 01:16:14,903
It's also then misunderstanding.

604
01:16:15,529 --> 01:16:17,281
This focused on the Nvidia stock.

605
01:16:17,281 --> 01:16:19,241
There's a misunderstanding of what the

606
01:16:19,241 --> 01:16:20,158
GPUs are actually

607
01:16:20,158 --> 01:16:21,451
exactly what their purpose is.

608
01:16:21,451 --> 01:16:24,413
Because yes, there's a lot of GPU, you

609
01:16:24,413 --> 01:16:25,622
know, and other compute

610
01:16:25,622 --> 01:16:27,124
power that is going into training

611
01:16:27,124 --> 01:16:27,708
these models.

612
01:16:28,125 --> 01:16:30,210
And so the argument is deep seek had a

613
01:16:30,210 --> 01:16:32,838
significantly less or

614
01:16:32,838 --> 01:16:34,965
smaller system set of

615
01:16:34,965 --> 01:16:35,173
wine.

616
01:16:35,507 --> 01:16:36,675
Less significantly less.

617
01:16:36,883 --> 01:16:38,176
Well, I think I'm going to I'm going to

618
01:16:38,176 --> 01:16:40,470
choose to believe right now because but

619
01:16:40,470 --> 01:16:41,138
that was just purely

620
01:16:41,138 --> 01:16:42,389
for the training aspect of thing.

621
01:16:42,431 --> 01:16:44,474
And then there's questions into maybe

622
01:16:44,474 --> 01:16:45,309
they were, you know, they

623
01:16:45,309 --> 01:16:47,603
jump a few leads ahead by

624
01:16:48,103 --> 01:16:49,146
potentially sitting on

625
01:16:49,146 --> 01:16:50,188
the shoulders of giants.

626
01:16:50,188 --> 01:16:51,273
So yeah, exactly right.

627
01:16:51,940 --> 01:16:53,734
But the most of those GPUs

628
01:16:53,734 --> 01:16:56,486
are not used for training.

629
01:16:56,486 --> 01:16:58,905
No, they used or at least you to type in

630
01:16:58,905 --> 01:17:00,324
going, hey, I need to do

631
01:17:00,324 --> 01:17:01,491
this or give me the answer for

632
01:17:01,491 --> 01:17:01,742
that.

633
01:17:01,742 --> 01:17:03,785
They can it is part of the inference, as

634
01:17:03,785 --> 01:17:04,953
you said, it is part of

635
01:17:04,953 --> 01:17:06,872
the output when the general

636
01:17:06,913 --> 01:17:08,332
users, businesses, you know,

637
01:17:08,582 --> 01:17:09,666
API calls that type of shit.

638
01:17:09,958 --> 01:17:11,251
It's them that with the computer.

639
01:17:11,501 --> 01:17:14,504
So yeah, Nvidia stocks crashing was an

640
01:17:14,504 --> 01:17:16,757
incorrect reaction or knee jerk reaction

641
01:17:16,757 --> 01:17:18,925
to this because deep

642
01:17:18,925 --> 01:17:20,677
seek is still going to need GPUs to

643
01:17:20,677 --> 01:17:22,679
delete a video for businesses to use

644
01:17:22,679 --> 01:17:25,390
their product on on scale or in

645
01:17:25,390 --> 01:17:25,724
scale.

646
01:17:26,016 --> 01:17:27,434
Exactly. And you know, like, as you said,

647
01:17:27,434 --> 01:17:28,560
there's a training side and

648
01:17:28,560 --> 01:17:29,603
then there's the inference side.

649
01:17:30,103 --> 01:17:33,523
And still needing a whole lot of this

650
01:17:33,523 --> 01:17:34,733
stuff for, you know, you

651
01:17:34,733 --> 01:17:36,443
still need the chips, you will get

652
01:17:36,443 --> 01:17:37,736
better results with better chips.

653
01:17:37,778 --> 01:17:39,112
So even though they say things like we

654
01:17:39,112 --> 01:17:40,280
didn't use the latest chips,

655
01:17:40,280 --> 01:17:41,698
well, yeah, you didn't need to

656
01:17:41,740 --> 01:17:42,574
exactly just to get to

657
01:17:42,574 --> 01:17:43,367
where you are right now.

658
01:17:43,742 --> 01:17:44,910
And imagine they did though.

659
01:17:45,160 --> 01:17:47,788
Oh, exactly. And they will like as things

660
01:17:47,788 --> 01:17:48,538
grow, it's not like

661
01:17:48,538 --> 01:17:49,956
they're just going to stop there.

662
01:17:49,956 --> 01:17:51,792
It's almost like people go, oh, they need

663
01:17:51,792 --> 01:17:52,709
like the latest chips.

664
01:17:53,085 --> 01:17:55,045
Well, they're going to, you know, this is

665
01:17:55,045 --> 01:17:56,296
the thing like inference is a

666
01:17:56,296 --> 01:17:58,548
much bigger thing in terms of

667
01:17:58,548 --> 01:17:59,966
usage because that's

668
01:17:59,966 --> 01:18:01,551
where normal people use it.

669
01:18:01,551 --> 01:18:03,220
And another thing that, you know, I often

670
01:18:03,220 --> 01:18:05,180
and you just remind me of, do you hear

671
01:18:05,180 --> 01:18:05,931
the myth that people

672
01:18:06,139 --> 01:18:07,891
think that as soon as you're using it,

673
01:18:08,183 --> 01:18:09,518
well, I can't type that in, it's going to

674
01:18:09,518 --> 01:18:10,310
train the model and I'll

675
01:18:10,310 --> 01:18:10,977
see it next time.

676
01:18:10,977 --> 01:18:11,645
And it's like, that's

677
01:18:11,645 --> 01:18:12,771
not how fast things work.

678
01:18:13,105 --> 01:18:15,107
People don't realize just how much you

679
01:18:15,107 --> 01:18:16,650
said training first and then there's

680
01:18:16,650 --> 01:18:17,526
inference in terms of

681
01:18:17,526 --> 01:18:19,903
usage. Training takes a long time.

682
01:18:19,903 --> 01:18:23,657
That's why like models that, for example,

683
01:18:23,990 --> 01:18:26,827
do graph and use graph like

684
01:18:26,827 --> 01:18:28,745
Neo4j and stuff like that to

685
01:18:28,829 --> 01:18:30,747
actually run rag because you get better

686
01:18:30,747 --> 01:18:32,124
connections between documents.

687
01:18:32,582 --> 01:18:34,042
That doesn't just happen straight away.

688
01:18:34,042 --> 01:18:36,294
And it's very costly because for them to

689
01:18:36,294 --> 01:18:39,131
retrain it, training is like a big thing.

690
01:18:39,131 --> 01:18:41,341
So you using it and, you know, putting in

691
01:18:41,341 --> 01:18:43,844
sensitive information isn't necessarily

692
01:18:43,844 --> 01:18:45,137
going to go into the model

693
01:18:45,178 --> 01:18:47,639
straight away. It may go into the next

694
01:18:47,639 --> 01:18:48,974
model when they train it.

695
01:18:48,974 --> 01:18:51,351
It may be used by researchers who can see

696
01:18:51,351 --> 01:18:53,437
that because you're putting data that

697
01:18:53,437 --> 01:18:54,563
goes into their database.

698
01:18:54,563 --> 01:18:56,440
OpenAI says that, you know, the data goes

699
01:18:56,440 --> 01:18:58,483
to the US, but it's not going to go

700
01:18:58,483 --> 01:18:59,443
straight into the model.

701
01:18:59,443 --> 01:19:00,402
Like you type something there.

702
01:19:00,402 --> 01:19:01,653
I'm not going to see it straight away.

703
01:19:01,695 --> 01:19:02,779
But people think that.

704
01:19:03,363 --> 01:19:04,739
So, you know, knowing the difference

705
01:19:04,739 --> 01:19:05,866
between training and

706
01:19:05,866 --> 01:19:07,200
inference, I think is key.

707
01:19:07,659 --> 01:19:08,452
But yeah, what else did

708
01:19:08,452 --> 01:19:09,786
you say about like deep seek?

709
01:19:10,829 --> 01:19:12,706
Well, I mean, the other talking about the

710
01:19:12,706 --> 01:19:14,166
impact is going to have on the market is

711
01:19:14,166 --> 01:19:15,500
for me, it's a positive one.

712
01:19:15,500 --> 01:19:17,043
I think at the first time, Altman kind

713
01:19:17,043 --> 01:19:21,173
of, you know, he kind of came out with a

714
01:19:21,173 --> 01:19:24,050
statement saying how, oh,

715
01:19:24,050 --> 01:19:25,385
this is a great model or this is going to

716
01:19:25,385 --> 01:19:25,927
put, you know, I

717
01:19:25,927 --> 01:19:26,720
should write an innovation.

718
01:19:26,928 --> 01:19:28,513
And then obviously soon after followed

719
01:19:28,513 --> 01:19:29,931
and started whinging about the fact that

720
01:19:29,931 --> 01:19:30,932
they were using their training

721
01:19:31,016 --> 01:19:32,976
data and how to let them

722
01:19:32,976 --> 01:19:34,394
cause the market crash, etc.

723
01:19:34,519 --> 01:19:37,063
They stole all of this training.

724
01:19:37,063 --> 01:19:38,106
Oh, God, where did you

725
01:19:38,106 --> 01:19:39,191
get your training data from?

726
01:19:40,358 --> 01:19:42,277
But it's open source.

727
01:19:42,736 --> 01:19:44,196
So not and this is going

728
01:19:44,196 --> 01:19:45,155
back on the Nvidia thing.

729
01:19:45,155 --> 01:19:47,324
So not only have we parked the inference

730
01:19:47,324 --> 01:19:52,662
side of things, businesses, startups who

731
01:19:52,662 --> 01:19:54,080
are now able to access this model to

732
01:19:54,080 --> 01:19:55,081
start doing their own,

733
01:19:55,624 --> 01:19:57,083
build their own AI businesses are still

734
01:19:57,083 --> 01:19:58,335
going to need this GPUs.

735
01:19:58,960 --> 01:20:01,588
So there's another added bonus to Nvidia

736
01:20:01,588 --> 01:20:04,049
now is like all this huge wave of other

737
01:20:04,049 --> 01:20:05,342
businesses and now is going to be able to

738
01:20:05,342 --> 01:20:07,010
get access to what will need access to

739
01:20:07,010 --> 01:20:08,345
all of these tools again.

740
01:20:09,304 --> 01:20:10,472
Not putting the need is massive

741
01:20:10,472 --> 01:20:11,640
opportunity, you know, the dip.

742
01:20:12,015 --> 01:20:14,601
Right. And yet what this because it is

743
01:20:14,601 --> 01:20:16,102
it's partial open source.

744
01:20:16,102 --> 01:20:18,271
They didn't release the data.

745
01:20:18,271 --> 01:20:18,897
I think they've released

746
01:20:18,897 --> 01:20:20,106
the weights or something.

747
01:20:20,106 --> 01:20:21,608
There's part because you need the data,

748
01:20:21,608 --> 01:20:24,319
but you also need the weights and they've

749
01:20:24,319 --> 01:20:25,362
released part of that.

750
01:20:25,362 --> 01:20:26,738
So it's not complete open source, but

751
01:20:26,738 --> 01:20:27,572
it's better than nothing.

752
01:20:27,614 --> 01:20:29,574
Right. And it means that businesses like

753
01:20:29,574 --> 01:20:31,409
and we've seen perplexity do it.

754
01:20:32,077 --> 01:20:33,537
Microsoft has taken a copy.

755
01:20:33,828 --> 01:20:34,287
You know what they call

756
01:20:34,287 --> 01:20:35,664
their thing? Think deeper.

757
01:20:35,997 --> 01:20:37,582
That is. Yes. So they've got their Think

758
01:20:37,582 --> 01:20:38,291
Deeper model, which

759
01:20:38,291 --> 01:20:40,377
runs the R1 DeepSeek stuff.

760
01:20:40,377 --> 01:20:41,503
You got something that I was going to say

761
01:20:41,503 --> 01:20:43,547
and perplexity integrated it as well.

762
01:20:43,547 --> 01:20:44,422
It's yes. Yeah.

763
01:20:44,422 --> 01:20:45,507
I am there at all. So you

764
01:20:45,507 --> 01:20:46,591
can actually go down here.

765
01:20:47,008 --> 01:20:49,386
So this leads to more people using it.

766
01:20:49,636 --> 01:20:51,221
And I think, you know, the open source

767
01:20:51,221 --> 01:20:53,265
now, you know, what you would have

768
01:20:53,265 --> 01:20:55,016
ignored because you can't fine tune.

769
01:20:55,725 --> 01:20:58,019
And I'll show you a Twitter thread where

770
01:20:58,019 --> 01:20:59,938
sorry, X thread where people are actually

771
01:20:59,938 --> 01:21:01,815
showing how you can fine tune your own.

772
01:21:01,815 --> 01:21:05,026
Right. So it is crazy what we are all of

773
01:21:05,026 --> 01:21:06,152
a sudden able to do.

774
01:21:06,152 --> 01:21:08,613
Like no one would have thought about, you

775
01:21:08,613 --> 01:21:10,865
know, building and doing their own kind

776
01:21:10,865 --> 01:21:12,158
of fine tuning to create their

777
01:21:12,158 --> 01:21:13,326
own model for their purposes.

778
01:21:13,618 --> 01:21:15,453
But now you can. Exactly.

779
01:21:15,495 --> 01:21:16,496
And show us this. So you

780
01:21:16,496 --> 01:21:17,414
see it here in perplexity.

781
01:21:17,414 --> 01:21:18,748
You just jumped out. You got promo.

782
01:21:19,124 --> 01:21:20,041
I've got promo. So then

783
01:21:20,041 --> 01:21:22,043
you got reasoning with R1.

784
01:21:22,377 --> 01:21:23,795
So it's integrated as well. So like you

785
01:21:23,795 --> 01:21:24,838
said, Google's done a bit of sync.

786
01:21:24,879 --> 01:21:25,630
That is really interesting. They have

787
01:21:25,630 --> 01:21:27,841
actually even which I'm surprised about

788
01:21:27,841 --> 01:21:29,259
because it's only come out within a day

789
01:21:29,259 --> 01:21:31,219
is I owe three mini models also there.

790
01:21:31,636 --> 01:21:32,971
Oh, yeah, that's interesting because then

791
01:21:32,971 --> 01:21:34,139
I just came out yesterday.

792
01:21:34,639 --> 01:21:36,308
Yeah. The heck I will show something on

793
01:21:36,308 --> 01:21:38,351
that because I've I've tested some stuff

794
01:21:38,351 --> 01:21:39,144
that I'll get to that.

795
01:21:39,144 --> 01:21:40,312
I think that that's out there.

796
01:21:41,229 --> 01:21:43,148
So, yeah, so that's and there's quite a

797
01:21:43,148 --> 01:21:43,857
few businesses that

798
01:21:43,857 --> 01:21:44,691
are starting to roll out.

799
01:21:44,691 --> 01:21:46,651
I said Microsoft, Exedy and others are

800
01:21:46,651 --> 01:21:47,944
coming out on many trying to test and

801
01:21:47,944 --> 01:21:49,988
learn Microsoft, even though they have,

802
01:21:51,323 --> 01:21:52,699
even though they are embracing it, they

803
01:21:52,699 --> 01:21:55,035
have also raised a few concerns around

804
01:21:55,035 --> 01:21:56,411
how the training of it

805
01:21:56,411 --> 01:21:57,495
has come into action.

806
01:21:57,495 --> 01:21:59,539
And they're digging into the code and

807
01:21:59,539 --> 01:22:00,790
going, oh, this is some things happening

808
01:22:00,790 --> 01:22:01,833
here and copied there.

809
01:22:01,833 --> 01:22:04,169
But like at the same time, it's still an

810
01:22:04,169 --> 01:22:06,129
awesome assistance for

811
01:22:06,129 --> 01:22:07,005
innovation within the market,

812
01:22:07,005 --> 01:22:08,798
because now it's going to put pressure on

813
01:22:08,798 --> 01:22:11,718
open AI to start properly thinking about

814
01:22:11,718 --> 01:22:14,721
how they do things in a more open way.

815
01:22:15,388 --> 01:22:15,847
Open way,

816
01:22:16,014 --> 01:22:17,349
potentially. Well, I don't know.

817
01:22:17,390 --> 01:22:18,683
I don't know about open way, but I guess

818
01:22:18,683 --> 01:22:20,352
like it's going to put pressure on them

819
01:22:20,352 --> 01:22:21,394
from an innovation standpoint.

820
01:22:21,895 --> 01:22:23,772
Yeah. And all the other businesses.

821
01:22:23,772 --> 01:22:25,607
Sam saying that he's going to you know,

822
01:22:25,607 --> 01:22:26,983
we've got some stuff up lastly,

823
01:22:26,983 --> 01:22:27,609
basically, has the

824
01:22:27,609 --> 01:22:28,943
effect of because of this.

825
01:22:29,152 --> 01:22:30,070
Like it's a great

826
01:22:30,070 --> 01:22:31,071
thing. Competition is great.

827
01:22:31,363 --> 01:22:33,281
I think that expedited the three mini

828
01:22:33,281 --> 01:22:35,867
launch that we've probably.

829
01:22:36,159 --> 01:22:38,787
No, no, exactly. Here's another example

830
01:22:38,787 --> 01:22:41,623
of so someone, Arby Chwala, that you

831
01:22:41,623 --> 01:22:42,707
might see on Twitter,

832
01:22:43,458 --> 01:22:44,668
he has shown how you

833
01:22:44,668 --> 01:22:46,127
can fine tune deep sea.

834
01:22:46,711 --> 01:22:47,671
So step by step, what

835
01:22:47,671 --> 01:22:50,173
you need to unsloth AI.

836
01:22:50,173 --> 01:22:51,800
That's a great Aussie company. I'm pretty

837
01:22:51,800 --> 01:22:52,425
sure they're Aussie.

838
01:22:53,468 --> 01:22:55,011
You need Olama. You know, in this

839
01:22:55,011 --> 01:22:57,055
example, how you load the model, then he

840
01:22:57,055 --> 01:22:59,057
defines how you do the LoRa config.

841
01:22:59,391 --> 01:23:00,141
So you change like a

842
01:23:00,141 --> 01:23:01,226
model weights and stuff.

843
01:23:01,226 --> 01:23:03,728
Then he gets preparing a dataset.

844
01:23:03,978 --> 01:23:05,313
He'll use the alpaca. There's a whole

845
01:23:05,313 --> 01:23:06,773
bunch of these different datasets online

846
01:23:06,773 --> 01:23:09,067
that you can use defining the

847
01:23:09,067 --> 01:23:10,860
trainer all the way to training

848
01:23:11,444 --> 01:23:13,279
and then exporting to Olama. And now it

849
01:23:13,279 --> 01:23:15,156
becomes a model that you have fine tuned.

850
01:23:15,532 --> 01:23:16,783
So I know I've just run through it

851
01:23:16,783 --> 01:23:18,118
quickly. And yet there's going to be some

852
01:23:18,118 --> 01:23:19,077
steps to figure it out.

853
01:23:19,077 --> 01:23:20,787
And sure, you're going to need to have

854
01:23:20,787 --> 01:23:22,372
machines available that can do this

855
01:23:22,372 --> 01:23:23,415
powerful enough to do this.

856
01:23:23,748 --> 01:23:26,084
But top end laptops or even just medium

857
01:23:26,084 --> 01:23:27,669
sized like laptops are going to be able

858
01:23:27,669 --> 01:23:28,795
to do this kind of stuff.

859
01:23:28,795 --> 01:23:32,173
Yeah, he's crazy. You know, and put it

860
01:23:32,173 --> 01:23:33,383
this way, talking about the opportunity.

861
01:23:34,175 --> 01:23:34,884
Still, there's going to be

862
01:23:34,884 --> 01:23:35,927
people that want to do it.

863
01:23:35,927 --> 01:23:37,178
But we're going to see this rise in

864
01:23:37,178 --> 01:23:38,763
businesses that become the middleware,

865
01:23:39,180 --> 01:23:40,807
that provide this service for you.

866
01:23:40,890 --> 01:23:42,350
OK, you want to fine tune. What do you

867
01:23:42,350 --> 01:23:43,727
want to fine tune? Here's the data.

868
01:23:44,102 --> 01:23:45,937
OK, it'll be back with you in a day.

869
01:23:46,604 --> 01:23:47,272
Click, click, click.

870
01:23:47,480 --> 01:23:48,732
Here you go. What the heck?

871
01:23:49,023 --> 01:23:51,401
Awesome. Yeah, it's great. So I'm big

872
01:23:51,401 --> 01:23:52,360
into this kind of stuff.

873
01:23:52,610 --> 01:23:53,945
Some of the some of the leaders in

874
01:23:53,945 --> 01:23:56,072
Australia, like across some of our

875
01:23:56,072 --> 01:23:58,742
biggest, you know, Australian companies,

876
01:23:58,742 --> 01:24:01,244
but companies. So Robin

877
01:24:01,244 --> 01:24:03,163
Kirk, a chunky chief executive.

878
01:24:03,496 --> 01:24:04,581
You know, the key thing is we're going to

879
01:24:04,581 --> 01:24:05,457
see efficiencies over

880
01:24:05,457 --> 01:24:07,333
time like any product.

881
01:24:08,793 --> 01:24:10,003
The Google Australia, New

882
01:24:10,003 --> 01:24:11,129
Zealand managing director.

883
01:24:11,671 --> 01:24:12,922
It's an endorsement for the impact of

884
01:24:12,922 --> 01:24:14,090
open source models,

885
01:24:15,049 --> 01:24:16,634
putting pressure on organization.

886
01:24:17,886 --> 01:24:19,888
It begs the question, the cost here, like

887
01:24:19,888 --> 01:24:21,306
Stephen Waller from Microsoft.

888
01:24:22,515 --> 01:24:24,142
You know, why are we spending these

889
01:24:24,142 --> 01:24:25,477
billions of dollars when you can train

890
01:24:25,477 --> 01:24:27,187
them for sets of millions?

891
01:24:27,937 --> 01:24:29,773
Christopher maneuver, twine chief

892
01:24:29,773 --> 01:24:32,192
technology officer, you know, this kind

893
01:24:32,192 --> 01:24:33,109
of transparency could

894
01:24:33,109 --> 01:24:34,277
really push things forward.

895
01:24:34,486 --> 01:24:35,612
Again, touching on some of the stuff

896
01:24:35,612 --> 01:24:36,780
that's probably changed in the US.

897
01:24:36,821 --> 01:24:38,740
But then also, you know, what is what's

898
01:24:38,740 --> 01:24:40,492
upcoming from open, et cetera.

899
01:24:41,201 --> 01:24:42,243
So just like some really interesting

900
01:24:42,243 --> 01:24:44,621
points of view around this, that it is

901
01:24:44,621 --> 01:24:45,538
going to put pressure,

902
01:24:45,538 --> 01:24:46,748
positive pressure on the market,

903
01:24:47,832 --> 01:24:49,542
although, you know, there are risks and

904
01:24:49,542 --> 01:24:50,001
there are some people

905
01:24:50,001 --> 01:24:51,002
do highlight those risks.

906
01:24:51,336 --> 01:24:53,963
It is a Chinese company.

907
01:24:54,380 --> 01:24:55,799
There is questions around what happens

908
01:24:55,799 --> 01:24:57,550
with your personal data.

909
01:24:57,550 --> 01:24:58,843
But if you sign up to it,

910
01:24:59,177 --> 01:25:00,261
that's a bit of an unknown yet.

911
01:25:00,261 --> 01:25:01,262
Obviously, if you run these models

912
01:25:01,262 --> 01:25:02,597
locally in your system, there's probably

913
01:25:02,597 --> 01:25:03,932
a reduction of risk there.

914
01:25:04,432 --> 01:25:06,059
But if you're using that, there are a lot

915
01:25:06,059 --> 01:25:07,811
of there's a basic question, what's

916
01:25:07,811 --> 01:25:09,896
happening with the data, et cetera.

917
01:25:09,896 --> 01:25:11,064
Yeah. So yeah, so

918
01:25:11,064 --> 01:25:11,648
it's pretty interesting.

919
01:25:11,856 --> 01:25:13,107
I'm pretty excited for it all,

920
01:25:13,441 --> 01:25:15,610
particularly the possibilities of even

921
01:25:15,610 --> 01:25:16,694
just like the running it

922
01:25:16,694 --> 01:25:17,737
locally on your own system.

923
01:25:18,071 --> 01:25:19,906
I haven't done it yet. We will.

924
01:25:19,906 --> 01:25:21,533
Apparently, it's not too difficult to do.

925
01:25:21,533 --> 01:25:25,870
Yeah. And and you're just you have this

926
01:25:25,870 --> 01:25:27,539
pressure of innovation on the market is

927
01:25:27,539 --> 01:25:29,123
just I'm here for it.

928
01:25:29,123 --> 01:25:29,499
That's what I want.

929
01:25:29,833 --> 01:25:31,626
Exactly. I'm here for it, too.

930
01:25:31,668 --> 01:25:33,336
And you just come back to and we'll cover

931
01:25:33,336 --> 01:25:35,547
off on like the questions and a few other

932
01:25:35,547 --> 01:25:37,048
things before we get into the other news.

933
01:25:37,048 --> 01:25:39,467
But, you know, folks are questioning one

934
01:25:39,467 --> 01:25:41,469
of the biggest things is like saying that

935
01:25:41,469 --> 01:25:42,262
they only trained for

936
01:25:42,262 --> 01:25:43,263
around six million dollars.

937
01:25:43,263 --> 01:25:44,931
Like, what does training actually mean?

938
01:25:45,056 --> 01:25:47,016
Kind of thing like they haven't built

939
01:25:47,016 --> 01:25:48,309
things from scratch.

940
01:25:48,810 --> 01:25:50,603
And some people are talking about like

941
01:25:50,603 --> 01:25:53,189
and maybe Daria Amadei, founder of

942
01:25:53,189 --> 01:25:54,816
Anthropic, has said this as well.

943
01:25:55,525 --> 01:25:57,652
But there's a whole lot of resource

944
01:25:57,652 --> 01:25:59,445
costs, you know, the compute, the

945
01:25:59,445 --> 01:26:00,655
resource, all the things

946
01:26:00,655 --> 01:26:01,406
that you need to build up

947
01:26:01,406 --> 01:26:04,450
to get to this point and just counting

948
01:26:04,450 --> 01:26:06,494
they believe because it's not I don't

949
01:26:06,494 --> 01:26:07,453
think deep sea because

950
01:26:07,453 --> 01:26:09,581
they put stuff out there,

951
01:26:09,581 --> 01:26:11,207
but there's no in-depth research that's

952
01:26:11,207 --> 01:26:13,042
not validated by others kind of research.

953
01:26:13,209 --> 01:26:15,461
So you have to take their word for it if

954
01:26:15,461 --> 01:26:16,296
you want to do that.

955
01:26:16,671 --> 01:26:18,882
But having something where you are

956
01:26:18,882 --> 01:26:20,550
building on top of others and only

957
01:26:20,550 --> 01:26:22,135
counting that, it's like, well,

958
01:26:22,135 --> 01:26:23,553
my coffee only cost me five dollars

959
01:26:23,553 --> 01:26:24,762
versus the guy down the road that

960
01:26:24,762 --> 01:26:26,890
actually grows the beans and, you know,

961
01:26:27,348 --> 01:26:29,350
harvests it himself, maybe puts it

962
01:26:29,350 --> 01:26:30,226
through what is it like?

963
01:26:30,226 --> 01:26:33,104
Hopi Luach goes through the Civit Cat in

964
01:26:33,104 --> 01:26:34,814
one and the Civit Cat out the other,

965
01:26:35,231 --> 01:26:36,733
clean it up again and go.

966
01:26:37,317 --> 01:26:39,277
And I get the world's most expensive

967
01:26:39,277 --> 01:26:41,404
coffee, for example, and pay 40 bucks for

968
01:26:41,404 --> 01:26:42,697
it, you know, versus the guy that

969
01:26:42,697 --> 01:26:44,198
actually raised the cat

970
01:26:45,575 --> 01:26:47,201
and the beans and did all that.

971
01:26:47,368 --> 01:26:49,203
So, you know, it's a matter of where do

972
01:26:49,203 --> 01:26:50,204
you measure and stuff.

973
01:26:50,204 --> 01:26:52,707
But the biggest point, I think, is the

974
01:26:52,707 --> 01:26:53,875
the opportunity and stuff.

975
01:26:53,875 --> 01:26:54,751
And that's what I said here.

976
01:26:55,668 --> 01:26:56,961
Here's this expert post.

977
01:26:57,545 --> 01:26:58,087
If you don't know this

978
01:26:58,087 --> 01:26:59,714
guy, folks, his name is Mark.

979
01:27:00,173 --> 01:27:01,424
He's friendly most of the time.

980
01:27:02,342 --> 01:27:06,179
He wrote this article about six key areas

981
01:27:06,179 --> 01:27:08,973
that builders should focus on.

982
01:27:09,390 --> 01:27:10,308
You know, this highlights

983
01:27:10,308 --> 01:27:11,476
that there's more constraints.

984
01:27:12,685 --> 01:27:15,730
When constraints are on you, it forces

985
01:27:15,730 --> 01:27:16,940
you to think outside the box.

986
01:27:17,231 --> 01:27:19,067
So constraint being that the latest chips

987
01:27:19,067 --> 01:27:19,859
aren't going to China.

988
01:27:20,276 --> 01:27:21,486
So that will force to, you

989
01:27:21,486 --> 01:27:23,279
know, think of clever ways.

990
01:27:23,321 --> 01:27:24,781
And, you know, Sam Altman and others

991
01:27:24,781 --> 01:27:26,824
acknowledge that this was a clever way to

992
01:27:26,824 --> 01:27:28,993
build the way that they took out the

993
01:27:28,993 --> 01:27:30,620
human feedback part of

994
01:27:30,620 --> 01:27:31,704
reinforcement learning.

995
01:27:31,704 --> 01:27:34,666
So R.L.H.F., which is what normally AI is

996
01:27:34,666 --> 01:27:38,086
trained on, having the machines actually

997
01:27:38,086 --> 01:27:40,088
reward itself is a big thing.

998
01:27:40,964 --> 01:27:42,548
The opportunity for middleware, like

999
01:27:42,548 --> 01:27:44,634
folks that provide tools that make these

1000
01:27:44,634 --> 01:27:47,845
services more accessible, that fed the

1001
01:27:47,845 --> 01:27:49,263
fuel of the fire kind of thing.

1002
01:27:50,056 --> 01:27:51,391
Data and governance are important.

1003
01:27:51,391 --> 01:27:53,059
If we can't trust the data, then, you

1004
01:27:53,059 --> 01:27:54,352
know, what does it matter with AI?

1005
01:27:54,936 --> 01:27:57,188
So, you know, having services out there

1006
01:27:57,188 --> 01:27:59,732
that do that sort of thing, integrators

1007
01:27:59,732 --> 01:28:00,900
being the key, like I said.

1008
01:28:01,526 --> 01:28:04,487
And it also highlights actually a big

1009
01:28:04,487 --> 01:28:05,655
thing around open source.

1010
01:28:06,030 --> 01:28:07,323
Just because something is open source

1011
01:28:07,323 --> 01:28:08,408
doesn't mean it's going to be great.

1012
01:28:08,658 --> 01:28:10,034
You have to have incentives, Chris.

1013
01:28:10,034 --> 01:28:10,576
So true.

1014
01:28:10,576 --> 01:28:12,996
You have to have ways to reward people to

1015
01:28:12,996 --> 01:28:15,164
keep contributing to open source.

1016
01:28:15,164 --> 01:28:16,457
Otherwise, if these things are just out

1017
01:28:16,457 --> 01:28:17,500
there, they'll just

1018
01:28:17,500 --> 01:28:19,168
decay if no one uses it.

1019
01:28:19,168 --> 01:28:20,962
Sure. Like we're all excited and stuff.

1020
01:28:21,170 --> 01:28:22,547
But if we want to have really good models

1021
01:28:22,547 --> 01:28:23,089
like this is an

1022
01:28:23,089 --> 01:28:25,550
opportunity for Australia, Mr.

1023
01:28:25,717 --> 01:28:28,136
Alvin Eazy, to have its own sovereign

1024
01:28:28,136 --> 01:28:30,013
kind of models like we can actually

1025
01:28:30,013 --> 01:28:31,514
better build these things.

1026
01:28:31,514 --> 01:28:34,142
The barrier is lower to actually do stuff

1027
01:28:34,142 --> 01:28:35,768
where we can protect ourselves just in

1028
01:28:35,768 --> 01:28:38,563
case, for example, there's a online

1029
01:28:38,563 --> 01:28:41,441
digital physical war that stops us having

1030
01:28:41,441 --> 01:28:43,484
access to because right now a lot of

1031
01:28:43,484 --> 01:28:45,945
Aussie businesses access overseas models.

1032
01:28:46,487 --> 01:28:48,114
Maybe it's like local because Microsoft

1033
01:28:48,114 --> 01:28:50,408
has a bit if something changes and we get

1034
01:28:50,408 --> 01:28:52,410
cut, you know, cut off from the rest of

1035
01:28:52,410 --> 01:28:54,078
the world, we can do better by

1036
01:28:54,078 --> 01:28:55,121
having our own kind of models.

1037
01:28:55,329 --> 01:28:56,748
So I think it's an interesting kind of

1038
01:28:56,748 --> 01:28:58,958
thought process from this expert that you

1039
01:28:58,958 --> 01:28:59,959
see on the screen right here.

1040
01:29:00,585 --> 01:29:01,669
Excellent. What do you

1041
01:29:01,669 --> 01:29:02,712
what do you got next?

1042
01:29:03,212 --> 01:29:04,172
Actually, you know,

1043
01:29:04,172 --> 01:29:05,131
should we go to the comparison?

1044
01:29:05,673 --> 01:29:06,841
Should we do a fun comparison or do you

1045
01:29:06,841 --> 01:29:07,884
want to do something else on deep?

1046
01:29:07,884 --> 01:29:08,968
Do you have deep six comparison?

1047
01:29:09,427 --> 01:29:10,261
I just do it.

1048
01:29:10,261 --> 01:29:10,887
Yeah, let's do it.

1049
01:29:11,012 --> 01:29:12,805
Oh, actually, no, no, not deep six.

1050
01:29:12,805 --> 01:29:14,724
Sorry, I'm thinking of oh, three.

1051
01:29:14,724 --> 01:29:15,683
I was going to say you go out.

1052
01:29:16,100 --> 01:29:17,185
Well, we'll jump to that later.

1053
01:29:17,185 --> 01:29:17,977
Let's go something else.

1054
01:29:17,977 --> 01:29:18,895
There's plenty of stuff out there.

1055
01:29:18,895 --> 01:29:19,854
People can look around and

1056
01:29:19,854 --> 01:29:20,938
check out what they've done.

1057
01:29:21,439 --> 01:29:23,566
But I mean, on the launch of new

1058
01:29:23,566 --> 01:29:27,320
platforms out there, we've I mean, out of

1059
01:29:27,320 --> 01:29:28,738
China is not the own deep sea is not the

1060
01:29:28,738 --> 01:29:30,865
only thing has come to what we got.

1061
01:29:30,865 --> 01:29:31,824
Come to the playing field.

1062
01:29:32,158 --> 01:29:36,579
Have we all heard of a good old Alibaba

1063
01:29:36,579 --> 01:29:39,874
and the Alibaba and they have released

1064
01:29:39,874 --> 01:29:41,667
something called Quinn.

1065
01:29:42,251 --> 01:29:43,211
Is it Alibaba in the

1066
01:29:43,211 --> 01:29:44,170
40 days or something?

1067
01:29:44,170 --> 01:29:44,670
What was the story?

1068
01:29:45,088 --> 01:29:46,422
Is that what I mean, Aladdin?

1069
01:29:47,840 --> 01:29:48,466
How about it?

1070
01:29:48,466 --> 01:29:50,343
Or is it a lot of the 40 for you to know

1071
01:29:50,343 --> 01:29:51,844
is Alibaba and the 40 days?

1072
01:29:52,178 --> 01:29:53,471
Yeah, I can't remember what was the

1073
01:29:53,471 --> 01:29:55,223
reasoning behind the name for

1074
01:29:55,223 --> 01:29:56,641
the Chinese company for that.

1075
01:29:56,641 --> 01:29:58,684
I have no idea that I was in correlation.

1076
01:29:58,684 --> 01:29:59,227
I'm going to look I'm

1077
01:29:59,227 --> 01:30:00,144
going to look it up anyway.

1078
01:30:00,144 --> 01:30:00,520
Keep going.

1079
01:30:00,853 --> 01:30:01,187
All right.

1080
01:30:01,187 --> 01:30:04,232
So when two point five slash Max has come

1081
01:30:04,232 --> 01:30:05,108
out, this is another

1082
01:30:05,108 --> 01:30:07,318
competitor entering the playing field.

1083
01:30:08,027 --> 01:30:08,945
I haven't had a big play

1084
01:30:08,945 --> 01:30:10,488
around with this at all.

1085
01:30:10,488 --> 01:30:11,364
I've been focused on

1086
01:30:11,364 --> 01:30:12,532
deep seek and open AI.

1087
01:30:13,407 --> 01:30:14,867
But they are.

1088
01:30:14,867 --> 01:30:16,494
Yeah, essentially they've come out saying

1089
01:30:16,494 --> 01:30:18,913
that they've got some incredible

1090
01:30:18,913 --> 01:30:19,997
performance metrics,

1091
01:30:20,998 --> 01:30:21,707
keeping up there like

1092
01:30:21,707 --> 01:30:22,917
better than the meta.

1093
01:30:23,209 --> 01:30:25,503
They're keeping up with open for a model.

1094
01:30:26,587 --> 01:30:30,216
And they have also got embedded into it

1095
01:30:30,216 --> 01:30:31,342
their image generation

1096
01:30:31,342 --> 01:30:32,802
and video generation.

1097
01:30:33,177 --> 01:30:34,220
Well, have you tried

1098
01:30:34,220 --> 01:30:34,846
the video generation?

1099
01:30:35,054 --> 01:30:35,721
I haven't found that I

1100
01:30:35,721 --> 01:30:36,681
find it's a bit slow.

1101
01:30:37,598 --> 01:30:39,600
When I did do a test at home.

1102
01:30:39,600 --> 01:30:41,102
Yeah, I got bored of waiting.

1103
01:30:41,227 --> 01:30:42,061
So I did stop.

1104
01:30:43,479 --> 01:30:44,856
But let's just do an image generation.

1105
01:30:45,940 --> 01:30:47,316
Generate an image.

1106
01:30:48,276 --> 01:30:49,360
With two folks host.

1107
01:30:49,610 --> 01:30:50,611
Two guys.

1108
01:30:51,279 --> 01:30:52,738
Yeah, two guys on a podcast.

1109
01:30:53,447 --> 01:30:54,490
No, don't do that.

1110
01:30:54,615 --> 01:30:55,116
No. OK.

1111
01:30:55,366 --> 01:30:57,201
And one says under fuel the Fed.

1112
01:30:57,994 --> 01:30:59,579
The under the hot sun.

1113
01:31:00,121 --> 01:31:00,329
OK.

1114
01:31:00,913 --> 01:31:01,539
Under the hot sun.

1115
01:31:01,831 --> 01:31:02,248
OK, fine.

1116
01:31:02,248 --> 01:31:02,456
I know.

1117
01:31:02,456 --> 01:31:02,707
Why not?

1118
01:31:03,833 --> 01:31:04,750
It takes a bit of time.

1119
01:31:05,168 --> 01:31:06,878
It's not it's not as fast as Dali.

1120
01:31:06,878 --> 01:31:08,129
But here we go.

1121
01:31:08,129 --> 01:31:08,671
Or 50 percent.

1122
01:31:08,671 --> 01:31:09,338
That's 60 percent.

1123
01:31:09,338 --> 01:31:10,089
It's a little bit quicker.

1124
01:31:10,089 --> 01:31:11,924
So I figured out what I figured out.

1125
01:31:11,924 --> 01:31:13,885
I used to figure out like why.

1126
01:31:14,385 --> 01:31:14,969
There we go.

1127
01:31:15,511 --> 01:31:15,720
Alibaba.

1128
01:31:15,845 --> 01:31:16,429
Oh, my gosh.

1129
01:31:16,679 --> 01:31:17,221
That's us.

1130
01:31:17,221 --> 01:31:17,930
So that's pretty good.

1131
01:31:18,181 --> 01:31:20,308
Like it's you know, it's a little bit

1132
01:31:20,308 --> 01:31:22,143
soft, like a little softer image.

1133
01:31:22,143 --> 01:31:23,394
I mean, they both need to brush their

1134
01:31:23,394 --> 01:31:25,188
teeth or teeth their brushes.

1135
01:31:25,188 --> 01:31:25,980
Chris might say.

1136
01:31:27,231 --> 01:31:27,982
But this guy looks like

1137
01:31:27,982 --> 01:31:29,192
he's got stuff in his teeth.

1138
01:31:29,567 --> 01:31:31,110
I think he was a former zombie.

1139
01:31:31,110 --> 01:31:31,527
Yangstra.

1140
01:31:31,819 --> 01:31:33,446
I mean, I just love that this this

1141
01:31:33,446 --> 01:31:34,989
podcast is attached to his.

1142
01:31:34,989 --> 01:31:35,364
This guy.

1143
01:31:35,740 --> 01:31:37,491
This mic is attached to his chest here.

1144
01:31:38,618 --> 01:31:39,118
This one.

1145
01:31:39,535 --> 01:31:40,620
This one's floating.

1146
01:31:40,620 --> 01:31:41,287
Let me try that.

1147
01:31:41,495 --> 01:31:41,704
OK.

1148
01:31:42,205 --> 01:31:43,789
Yeah, that I mean, look, this is the look

1149
01:31:43,789 --> 01:31:44,749
for twenty twenty five.

1150
01:31:44,749 --> 01:31:47,543
But like the hands.

1151
01:31:47,543 --> 01:31:48,544
That is a good image.

1152
01:31:48,544 --> 01:31:49,712
The hands are right.

1153
01:31:49,712 --> 01:31:50,171
Except that.

1154
01:31:50,171 --> 01:31:51,505
No, no, that hand there.

1155
01:31:51,505 --> 01:31:53,007
His thumb is like really.

1156
01:31:53,007 --> 01:31:53,925
Oh, yeah, it's a little bit.

1157
01:31:53,925 --> 01:31:55,426
But I mean, you could it's possible.

1158
01:31:55,426 --> 01:31:56,219
Could have been an accident.

1159
01:31:57,178 --> 01:31:57,887
But you mean I've also

1160
01:31:57,887 --> 01:31:58,679
just put one problem.

1161
01:31:58,971 --> 01:31:59,096
Wow.

1162
01:31:59,096 --> 01:32:01,390
And and then if I if I go for videos.

1163
01:32:01,724 --> 01:32:02,558
Oh, it's coming soon.

1164
01:32:02,850 --> 01:32:03,434
Oh, it was great.

1165
01:32:03,643 --> 01:32:04,018
I know.

1166
01:32:04,018 --> 01:32:04,685
Put the prompt in.

1167
01:32:05,645 --> 01:32:07,271
No, no, it's almost there.

1168
01:32:07,563 --> 01:32:09,023
OK, but it was available for me.

1169
01:32:09,482 --> 01:32:09,857
That's weird.

1170
01:32:10,107 --> 01:32:10,399
All right.

1171
01:32:10,399 --> 01:32:10,858
OK, fine.

1172
01:32:10,858 --> 01:32:13,778
They'll find Chris a fan of them so much.

1173
01:32:13,778 --> 01:32:15,780
But the reason why I was named Alibaba

1174
01:32:15,780 --> 01:32:18,032
was because of various things,

1175
01:32:18,032 --> 01:32:20,326
the easy to pronounce, but also the open

1176
01:32:20,326 --> 01:32:21,619
sesame kind of connection,

1177
01:32:21,911 --> 01:32:24,997
unlocking a secret treasure and symbolism

1178
01:32:24,997 --> 01:32:26,916
as well of wealth and opportunity.

1179
01:32:27,375 --> 01:32:28,501
So there you go.

1180
01:32:29,377 --> 01:32:31,587
Now, obviously, it could be who's saying

1181
01:32:31,587 --> 01:32:33,172
that that, you know, that's

1182
01:32:33,172 --> 01:32:34,674
that's something to check.

1183
01:32:35,341 --> 01:32:37,009
Pretty sure it's not open source, though.

1184
01:32:37,301 --> 01:32:38,970
No, no, not yet.

1185
01:32:39,345 --> 01:32:40,513
It's yeah, it's competing.

1186
01:32:40,930 --> 01:32:42,348
You know, it's not as good as the others,

1187
01:32:42,348 --> 01:32:43,182
I would say right now.

1188
01:32:43,182 --> 01:32:44,100
It's it's got a it's

1189
01:32:44,100 --> 01:32:45,017
got a bit to catch up.

1190
01:32:45,268 --> 01:32:46,519
A little bit, a little bit.

1191
01:32:46,811 --> 01:32:48,020
But and more interesting news.

1192
01:32:48,646 --> 01:32:51,023
Oh, three mini is a more or three.

1193
01:32:51,023 --> 01:32:52,233
Oh, three. I mean, it is.

1194
01:32:52,233 --> 01:32:52,775
Oh, three mini.

1195
01:32:52,775 --> 01:32:54,360
Usually most of the three coming.

1196
01:32:54,860 --> 01:32:56,237
This would be the thing that is the

1197
01:32:56,237 --> 01:32:57,238
biggest bit of news.

1198
01:32:57,363 --> 01:32:58,281
Yeah, is that this week.

1199
01:32:58,281 --> 01:32:59,949
And we're relegating it.

1200
01:32:59,949 --> 01:33:00,533
Sorry, Sam.

1201
01:33:00,533 --> 01:33:01,701
I mean, come out with more stuff.

1202
01:33:01,701 --> 01:33:03,369
But like this is being pushed down

1203
01:33:03,369 --> 01:33:05,746
because it had such an

1204
01:33:05,746 --> 01:33:07,456
impact on the global.

1205
01:33:07,498 --> 01:33:09,792
I mean, this three didn't wipe out

1206
01:33:09,792 --> 01:33:11,794
trillions of dollars or billions of

1207
01:33:11,794 --> 01:33:13,587
dollars, sorry, from from Nvidia.

1208
01:33:14,046 --> 01:33:16,590
I mean, if it does, what a great buying

1209
01:33:16,590 --> 01:33:17,550
opportunity, not financial

1210
01:33:17,550 --> 01:33:19,594
advice, but they release this.

1211
01:33:19,844 --> 01:33:20,845
So if you are in chat,

1212
01:33:20,845 --> 01:33:23,556
you can select from it now.

1213
01:33:23,806 --> 01:33:25,391
They've released it on the free version

1214
01:33:25,391 --> 01:33:26,434
because there are people out there that

1215
01:33:26,434 --> 01:33:27,435
use the free version and

1216
01:33:27,435 --> 01:33:28,185
it's going to be limited.

1217
01:33:29,020 --> 01:33:30,855
And it's clever in that if you keep on

1218
01:33:30,855 --> 01:33:32,398
chatting with the free version, it asks,

1219
01:33:32,606 --> 01:33:33,482
don't you just want to log in?

1220
01:33:33,983 --> 01:33:36,027
Wouldn't you just like want to log in and

1221
01:33:36,027 --> 01:33:37,069
come in and like save

1222
01:33:37,069 --> 01:33:37,945
your chats and stuff?

1223
01:33:37,945 --> 01:33:41,073
Yep. But also for the API and what I

1224
01:33:41,073 --> 01:33:42,325
thought would be cool.

1225
01:33:43,451 --> 01:33:46,203
So I built a little thing just to test.

1226
01:33:46,704 --> 01:33:48,039
OK, so we've got three here.

1227
01:33:48,748 --> 01:33:49,248
We've got some of the

1228
01:33:49,248 --> 01:33:49,999
other different models.

1229
01:33:49,999 --> 01:33:50,875
I can put some other ones up

1230
01:33:50,875 --> 01:33:52,376
there, but you can see here.

1231
01:33:54,045 --> 01:33:57,882
Tell me how to tie my

1232
01:33:57,882 --> 01:34:01,218
shoes, but do it with math code.

1233
01:34:01,510 --> 01:34:03,387
I don't know. For example, do it.

1234
01:34:03,429 --> 01:34:04,889
Do it in code. Do it in

1235
01:34:04,889 --> 01:34:07,475
code for a robot, for example.

1236
01:34:07,808 --> 01:34:10,686
And so three doesn't do streaming text.

1237
01:34:11,312 --> 01:34:12,938
I think I tried that and said that the

1238
01:34:12,938 --> 01:34:14,607
model wouldn't let you do that yet.

1239
01:34:15,816 --> 01:34:16,567
And streaming. What do

1240
01:34:16,567 --> 01:34:17,151
you mean by streaming?

1241
01:34:17,151 --> 01:34:19,278
As in like it'll just stream it out

1242
01:34:19,278 --> 01:34:20,488
rather than the answer

1243
01:34:20,488 --> 01:34:21,405
comes out all at once.

1244
01:34:21,405 --> 01:34:23,240
But you go here's a bit of pseudo code

1245
01:34:23,240 --> 01:34:24,575
for a robot tying a shoe.

1246
01:34:24,825 --> 01:34:26,327
Cool. But what I wanted to do, an

1247
01:34:26,327 --> 01:34:27,495
explanation of the steps.

1248
01:34:27,953 --> 01:34:29,205
OK, so it does its thinking.

1249
01:34:29,413 --> 01:34:31,499
You see the thinking when you're on on

1250
01:34:31,499 --> 01:34:32,792
the chat GBT version.

1251
01:34:33,250 --> 01:34:34,627
What I thought I'd do is just do a little

1252
01:34:34,627 --> 01:34:36,796
comparison where I can pick

1253
01:34:36,796 --> 01:34:37,755
a couple of different tools.

1254
01:34:37,755 --> 01:34:38,881
So here's O3 Mini, which

1255
01:34:38,881 --> 01:34:40,049
is available by the API.

1256
01:34:40,633 --> 01:34:42,301
O1, the main model.

1257
01:34:43,552 --> 01:34:45,262
And as we were discussing before the

1258
01:34:45,262 --> 01:34:46,972
show, there are different

1259
01:34:46,972 --> 01:34:48,808
purposes for different models.

1260
01:34:49,183 --> 01:34:51,102
And so one model may be really good at

1261
01:34:51,102 --> 01:34:52,895
coding, but not good at storytelling.

1262
01:34:53,229 --> 01:34:54,939
One model may be really good at like, you

1263
01:34:54,939 --> 01:34:56,190
know, other things out there.

1264
01:34:56,440 --> 01:34:58,359
So and this is where O3 comes into play.

1265
01:34:58,401 --> 01:35:01,320
It is being promoted as that sort of code

1266
01:35:01,320 --> 01:35:05,366
slash STEM, so education and kind of

1267
01:35:05,366 --> 01:35:06,617
researchy type stuff.

1268
01:35:06,617 --> 01:35:09,537
So exactly where you might use for for

1269
01:35:09,537 --> 01:35:12,957
more, you know, that creative maybe

1270
01:35:12,957 --> 01:35:15,292
research and, you know,

1271
01:35:15,292 --> 01:35:17,044
just general conversations you would

1272
01:35:17,044 --> 01:35:19,213
switch to and you might switch to O1 for

1273
01:35:19,213 --> 01:35:20,005
something a bit more in

1274
01:35:20,005 --> 01:35:20,923
depth on the research side.

1275
01:35:20,923 --> 01:35:22,425
I think it was you'll use O3 for that

1276
01:35:22,425 --> 01:35:23,092
code generation,

1277
01:35:23,092 --> 01:35:25,094
education, engagement of content.

1278
01:35:25,136 --> 01:35:29,598
And it only recently the GPT40 Mini was

1279
01:35:29,598 --> 01:35:31,308
the one that was like touted

1280
01:35:31,308 --> 01:35:33,644
as the best coding out of like,

1281
01:35:33,811 --> 01:35:35,312
for O versus 40 Mini.

1282
01:35:35,563 --> 01:35:37,231
Mini was much more trained on coding.

1283
01:35:37,440 --> 01:35:38,190
So you use that for coding.

1284
01:35:38,524 --> 01:35:41,318
But in any case, I've this is a not so

1285
01:35:41,318 --> 01:35:42,361
like comparison, but

1286
01:35:42,361 --> 01:35:43,362
just interesting to see.

1287
01:35:43,696 --> 01:35:45,156
I said, tell me a joke about OpenAI

1288
01:35:45,156 --> 01:35:46,407
complaining about the theft of their

1289
01:35:46,407 --> 01:35:47,450
model when they copied a

1290
01:35:47,450 --> 01:35:48,367
lot of everyone's data,

1291
01:35:48,659 --> 01:35:49,910
which is obviously they

1292
01:35:49,910 --> 01:35:51,120
stole my drugs kind of moment.

1293
01:35:51,620 --> 01:35:53,372
O3 Mini says, here's one.

1294
01:35:53,831 --> 01:35:55,291
OpenAI was fearing that someone stealing

1295
01:35:55,291 --> 01:35:56,709
their model, their spokesperson said,

1296
01:35:56,709 --> 01:35:57,418
we're not really mad.

1297
01:35:57,418 --> 01:35:59,044
We just realized we already threw a

1298
01:35:59,044 --> 01:36:00,504
potluck with everyone's data.

1299
01:36:00,713 --> 01:36:02,465
Looks like our work was a

1300
01:36:02,465 --> 01:36:03,757
group project from the start.

1301
01:36:04,258 --> 01:36:05,468
OK, kind of funny.

1302
01:36:05,885 --> 01:36:06,469
A little satire.

1303
01:36:07,344 --> 01:36:09,722
O1 said, OpenAI was furious someone

1304
01:36:09,722 --> 01:36:10,806
copied their model code.

1305
01:36:10,806 --> 01:36:12,558
Then someone asked, didn't you use all

1306
01:36:12,558 --> 01:36:13,851
our public data for training to which

1307
01:36:13,851 --> 01:36:15,728
OpenAI replied, shh, that's classified.

1308
01:36:15,978 --> 01:36:16,854
But hands off our stuff.

1309
01:36:17,146 --> 01:36:18,022
That's funnier.

1310
01:36:18,647 --> 01:36:19,231
I would say it's funnier.

1311
01:36:20,232 --> 01:36:22,860
4.0 Mini said, why did OpenAI start a

1312
01:36:22,860 --> 01:36:23,944
support group for their models?

1313
01:36:23,944 --> 01:36:25,321
Because they felt a little stolen.

1314
01:36:25,321 --> 01:36:26,363
They realized they had borrowed so much

1315
01:36:26,363 --> 01:36:27,198
data, they couldn't remember

1316
01:36:27,198 --> 01:36:28,365
who's it was in the first place.

1317
01:36:28,657 --> 01:36:32,912
And then 4.0 said, why did OpenAI call

1318
01:36:32,912 --> 01:36:34,246
the police about their stolen model?

1319
01:36:34,622 --> 01:36:36,207
Because they finally found something they

1320
01:36:36,207 --> 01:36:37,166
didn't copy from everyone.

1321
01:36:38,125 --> 01:36:39,585
That's probably the best one.

1322
01:36:40,461 --> 01:36:41,545
Human, human rating.

1323
01:36:41,795 --> 01:36:43,339
I think you can see which one we like.

1324
01:36:43,797 --> 01:36:47,426
Now, I've asked 4.0 Mini and 4.0 Mini

1325
01:36:47,426 --> 01:36:48,844
just, I mean, I could have

1326
01:36:48,844 --> 01:36:50,137
asked everyone to grade it.

1327
01:36:50,179 --> 01:36:51,722
But let's just ask for a

1328
01:36:51,722 --> 01:36:52,806
Mini to grade the response.

1329
01:36:53,140 --> 01:36:55,476
So, you know, here's something that we

1330
01:36:55,476 --> 01:36:56,769
just cooked up last night.

1331
01:36:56,769 --> 01:36:58,229
So it'll take a few seconds there.

1332
01:36:58,229 --> 01:37:00,356
But let's see what the AI actually thinks

1333
01:37:00,356 --> 01:37:02,107
is the best out of like

1334
01:37:02,107 --> 01:37:03,067
all the models and stuff.

1335
01:37:03,526 --> 01:37:04,610
So, oh, interesting.

1336
01:37:05,402 --> 01:37:05,819
That's weird.

1337
01:37:06,195 --> 01:37:06,987
I mean, it loves it.

1338
01:37:06,987 --> 01:37:07,655
Loves the detail and

1339
01:37:07,655 --> 01:37:08,531
description of the data.

1340
01:37:08,822 --> 01:37:10,824
That's why it's the firm, the three.

1341
01:37:10,824 --> 01:37:12,743
We could test this with coding problems.

1342
01:37:12,743 --> 01:37:14,286
We could test this with other ones.

1343
01:37:14,286 --> 01:37:15,037
And so this was like

1344
01:37:15,037 --> 01:37:16,622
a joke on creativity.

1345
01:37:17,081 --> 01:37:17,957
You deal with coding, you

1346
01:37:17,957 --> 01:37:18,958
might get different results.

1347
01:37:18,958 --> 01:37:19,959
And as we said, they're

1348
01:37:19,959 --> 01:37:21,126
all for different things.

1349
01:37:21,585 --> 01:37:23,337
But yeah, what do you think then?

1350
01:37:23,754 --> 01:37:26,674
I think that I don't

1351
01:37:26,674 --> 01:37:27,800
agree with those responses.

1352
01:37:28,092 --> 01:37:28,801
But you know, it's fun.

1353
01:37:29,218 --> 01:37:30,177
Do we know what we liked, right?

1354
01:37:30,177 --> 01:37:31,095
We know what we liked.

1355
01:37:31,095 --> 01:37:32,054
We wanted the jokes.

1356
01:37:32,054 --> 01:37:33,514
We like 4.0 Mini.

1357
01:37:33,722 --> 01:37:34,807
Like those are the good ones.

1358
01:37:34,807 --> 01:37:35,599
Like we're there for the jokes.

1359
01:37:35,599 --> 01:37:36,559
But yeah, you can definitely do some

1360
01:37:36,559 --> 01:37:37,351
other things with this.

1361
01:37:37,768 --> 01:37:38,686
This is just a little test.

1362
01:37:38,978 --> 01:37:40,271
The other thing to think about if we take

1363
01:37:40,271 --> 01:37:41,605
it from a human response point of view is

1364
01:37:41,605 --> 01:37:43,274
that you also are putting in a query

1365
01:37:43,274 --> 01:37:45,192
there that isn't

1366
01:37:45,192 --> 01:37:46,860
tailored to what the model is

1367
01:37:48,320 --> 01:37:50,197
stouted as being the best at.

1368
01:37:50,447 --> 01:37:50,614
Yeah.

1369
01:37:50,781 --> 01:37:52,241
If we start switching things to like

1370
01:37:52,241 --> 01:37:54,577
interesting like code, for example, we

1371
01:37:54,577 --> 01:37:57,538
might get some different responses where

1372
01:37:57,538 --> 01:37:58,289
we would put

1373
01:37:58,289 --> 01:37:59,665
favoritism towards the O3 model.

1374
01:37:59,915 --> 01:38:01,750
A lot of people do the code testing like

1375
01:38:01,750 --> 01:38:04,086
build me a game of Snake and they one

1376
01:38:04,086 --> 01:38:06,130
shot it and see which one will actually

1377
01:38:06,130 --> 01:38:07,172
build the game of Snake.

1378
01:38:07,506 --> 01:38:08,674
Well, I don't have that here.

1379
01:38:08,674 --> 01:38:09,675
I didn't have enough time.

1380
01:38:09,675 --> 01:38:11,844
But here is the coding example, O3 Mini

1381
01:38:11,844 --> 01:38:14,221
does for code versus like the others.

1382
01:38:14,221 --> 01:38:16,432
And then I can grade the responses.

1383
01:38:16,682 --> 01:38:17,641
It's interesting you said that.

1384
01:38:17,641 --> 01:38:20,728
So this post came out from a K.

1385
01:38:22,104 --> 01:38:24,356
So open O3 just one shot at this.

1386
01:38:24,356 --> 01:38:25,691
So they did a prompt where they write a

1387
01:38:25,691 --> 01:38:27,359
script of a hundred bouncing yellow balls

1388
01:38:27,359 --> 01:38:29,528
within a sphere to make sure it handles

1389
01:38:29,528 --> 01:38:30,863
the collision detection properly.

1390
01:38:31,113 --> 01:38:32,698
Make the sphere slowly rotate.

1391
01:38:32,698 --> 01:38:34,408
Make sure the ball stay within the sphere

1392
01:38:34,408 --> 01:38:36,952
and implement it in P5.

1393
01:38:36,952 --> 01:38:37,202
Yes.

1394
01:38:37,870 --> 01:38:38,078
Wow.

1395
01:38:38,996 --> 01:38:41,749
And I can just play it.

1396
01:38:41,999 --> 01:38:42,541
I want to go on a

1397
01:38:42,541 --> 01:38:43,667
full screen for you all.

1398
01:38:44,209 --> 01:38:44,835
And you can see there

1399
01:38:44,835 --> 01:38:46,211
that it is generating.

1400
01:38:46,754 --> 01:38:47,463
That's ridiculous.

1401
01:38:47,838 --> 01:38:49,715
The generating that model, that 3D model,

1402
01:38:49,923 --> 01:38:52,551
which is obviously done using Python and

1403
01:38:52,551 --> 01:38:53,469
math and other things.

1404
01:38:53,802 --> 01:38:57,598
But it's done it really, really cleanly.

1405
01:38:57,598 --> 01:38:58,182
That's spectacular.

1406
01:38:58,432 --> 01:38:59,308
One shot at it.

1407
01:38:59,642 --> 01:39:00,059
So unbelievable.

1408
01:39:00,726 --> 01:39:01,644
You know, the fact that you can do things

1409
01:39:01,644 --> 01:39:04,146
like that would quite complex algorithms

1410
01:39:04,146 --> 01:39:06,440
and spitting out something that accurate

1411
01:39:06,440 --> 01:39:07,900
is is quite impressive.

1412
01:39:08,484 --> 01:39:10,069
And just quickly, I just asked to do that

1413
01:39:10,069 --> 01:39:12,946
code thing and O3 Mini does win in terms

1414
01:39:12,946 --> 01:39:14,073
of that code that was

1415
01:39:14,073 --> 01:39:15,699
generated and getting,

1416
01:39:16,158 --> 01:39:18,827
you know, the AI to say why it is.

1417
01:39:19,203 --> 01:39:19,787
Sure, it's AI.

1418
01:39:20,245 --> 01:39:20,996
Judge it for yourself.

1419
01:39:20,996 --> 01:39:21,830
Judge it by the results.

1420
01:39:21,830 --> 01:39:23,207
Actually run it like you saw there.

1421
01:39:23,540 --> 01:39:25,542
But really interesting to, you know, when

1422
01:39:25,542 --> 01:39:28,253
you and what you'd want to do, probably

1423
01:39:28,253 --> 01:39:29,922
be more scientific with this and maybe

1424
01:39:29,922 --> 01:39:30,631
you do it yourself

1425
01:39:30,631 --> 01:39:31,924
manually like I'm doing now.

1426
01:39:31,924 --> 01:39:34,176
You get the AI to actually run a hundred

1427
01:39:34,176 --> 01:39:36,553
simulations or whatever to see like what

1428
01:39:36,553 --> 01:39:38,097
is the best result, I think

1429
01:39:38,097 --> 01:39:39,598
is definitely interesting.

1430
01:39:40,599 --> 01:39:41,809
But there were other things in the news.

1431
01:39:42,226 --> 01:39:44,353
Chris, I wanted to jump

1432
01:39:44,353 --> 01:39:45,604
to something like, say.

1433
01:39:46,397 --> 01:39:47,022
Operator.

1434
01:39:48,607 --> 01:39:48,816
Operator.

1435
01:39:49,358 --> 01:39:50,859
Lord and Sabia, Sam.

1436
01:39:51,276 --> 01:39:52,528
I haven't had a chance to play with this

1437
01:39:52,528 --> 01:39:53,445
because technically it's

1438
01:39:53,445 --> 01:39:54,363
not available in Australia.

1439
01:39:54,613 --> 01:39:55,030
Technically.

1440
01:39:55,364 --> 01:39:56,782
I mean, you don't want to pay three

1441
01:39:56,782 --> 01:39:58,033
hundred dollars Aussie a month.

1442
01:39:58,033 --> 01:39:58,742
I don't know.

1443
01:39:58,742 --> 01:39:59,493
I do not either.

1444
01:40:00,786 --> 01:40:02,246
But I think it is a big thing and it is

1445
01:40:02,246 --> 01:40:04,540
very exciting of what is to come.

1446
01:40:04,998 --> 01:40:05,624
Yeah, exactly.

1447
01:40:05,666 --> 01:40:08,544
There's a couple of examples here being

1448
01:40:08,544 --> 01:40:10,796
shown by what it can do, like searching

1449
01:40:10,796 --> 01:40:14,216
the web and performing tasks as Shruti

1450
01:40:14,216 --> 01:40:16,009
Mishra is showcasing here.

1451
01:40:16,301 --> 01:40:16,677
Crypto investment

1452
01:40:16,677 --> 01:40:18,846
research based on tokens.

1453
01:40:19,179 --> 01:40:20,389
So like here is some

1454
01:40:20,389 --> 01:40:22,766
research that it's doing on screen.

1455
01:40:24,184 --> 01:40:26,061
It's searching that you can actually see

1456
01:40:26,061 --> 01:40:28,230
it opening up a browser, running a

1457
01:40:28,230 --> 01:40:31,316
search, clicking on different things.

1458
01:40:32,317 --> 01:40:33,944
And it's kind of funny, it's kind of like

1459
01:40:33,944 --> 01:40:35,863
a really accurate, you know, when you

1460
01:40:35,863 --> 01:40:37,448
move the mouse, it's moving the mouse

1461
01:40:37,448 --> 01:40:38,282
here, but it's going

1462
01:40:38,282 --> 01:40:39,867
directly to things and stuff.

1463
01:40:40,117 --> 01:40:41,577
So it'll open up a few bits

1464
01:40:41,577 --> 01:40:43,579
of research there for you.

1465
01:40:43,579 --> 01:40:45,414
That is one that you can get it to do.

1466
01:40:45,414 --> 01:40:46,582
You can ask it to buy you

1467
01:40:46,582 --> 01:40:48,417
an iced coffee on Uber Eats.

1468
01:40:48,959 --> 01:40:50,169
So that's what this person is doing.

1469
01:40:50,419 --> 01:40:53,088
I probably I mean, look, I'd rather just

1470
01:40:53,088 --> 01:40:54,548
use the Uber Eats app, I think.

1471
01:40:54,798 --> 01:40:55,966
Right. Like the control operator.

1472
01:40:56,383 --> 01:40:58,469
It's interesting because it gets you to

1473
01:40:58,469 --> 01:41:01,180
log in and, you know,

1474
01:41:01,638 --> 01:41:02,931
it's not using your browser.

1475
01:41:02,931 --> 01:41:04,266
It's actually using a

1476
01:41:04,266 --> 01:41:05,100
what do you call it?

1477
01:41:05,100 --> 01:41:08,020
A virtual browser where it

1478
01:41:08,020 --> 01:41:10,022
does have to ask you to log in.

1479
01:41:10,022 --> 01:41:11,940
So there's an element of trust here,

1480
01:41:11,940 --> 01:41:13,525
Chris, like with operator in terms of

1481
01:41:13,525 --> 01:41:14,318
what you want it to do.

1482
01:41:14,693 --> 01:41:16,195
Claude came out with something a few

1483
01:41:16,195 --> 01:41:18,489
months ago, remember, where it was like

1484
01:41:18,489 --> 01:41:20,240
doing this kind of stuff and you could

1485
01:41:20,240 --> 01:41:21,742
install it and run it.

1486
01:41:21,784 --> 01:41:24,369
So, yes, some other interesting things.

1487
01:41:24,578 --> 01:41:25,746
Automating tweets is what

1488
01:41:25,746 --> 01:41:26,997
some people are using it for.

1489
01:41:27,706 --> 01:41:29,124
Finding the best and cheapest health

1490
01:41:29,124 --> 01:41:30,542
insurance in Switzerland, dinner

1491
01:41:30,542 --> 01:41:31,710
reservations again,

1492
01:41:31,710 --> 01:41:32,669
planning weekend trips,

1493
01:41:33,420 --> 01:41:35,380
finding the best route from Marrakesh to

1494
01:41:35,380 --> 01:41:36,924
Casablanca, you know, normal everyday

1495
01:41:36,924 --> 01:41:39,134
stuff that we see here in Sydney.

1496
01:41:40,093 --> 01:41:41,053
So, I mean, that's

1497
01:41:41,053 --> 01:41:42,471
that's one thing there.

1498
01:41:42,471 --> 01:41:44,223
And then other

1499
01:41:44,223 --> 01:41:45,891
examples here from operators.

1500
01:41:46,350 --> 01:41:47,976
So someone is getting to act as a

1501
01:41:47,976 --> 01:41:50,687
sorcerer and how it will impact HR.

1502
01:41:51,021 --> 01:41:54,399
So doing search on people, I think via

1503
01:41:54,399 --> 01:41:57,736
LinkedIn and seeing what it is that, you

1504
01:41:57,736 --> 01:41:58,946
know, the candidates have.

1505
01:41:58,946 --> 01:41:59,696
And evaluating them.

1506
01:41:59,947 --> 01:42:01,615
If obviously I didn't give a great

1507
01:42:01,615 --> 01:42:05,744
criteria to this, but no, this is.

1508
01:42:07,788 --> 01:42:08,121
He's shocked.

1509
01:42:09,414 --> 01:42:09,873
He is shocked.

1510
01:42:11,875 --> 01:42:12,292
Yeah.

1511
01:42:21,134 --> 01:42:22,594
Yeah, I mean, like definitely like tools

1512
01:42:22,594 --> 01:42:24,555
like this or just even chat GPT when it

1513
01:42:24,555 --> 01:42:26,223
came out in November 2022, people were

1514
01:42:26,223 --> 01:42:27,432
just shocked by it and

1515
01:42:27,432 --> 01:42:29,810
seeing the magic that it was.

1516
01:42:30,018 --> 01:42:33,647
It's pretty crazy if when I mean, I'm, I

1517
01:42:33,647 --> 01:42:36,608
hope that they release something for us

1518
01:42:36,608 --> 01:42:37,985
little plus users, which I

1519
01:42:37,985 --> 01:42:39,111
think they say he hinted at.

1520
01:42:39,111 --> 01:42:40,821
Plus you will get a hundred.

1521
01:42:41,071 --> 01:42:41,738
A hundred tokens.

1522
01:42:42,114 --> 01:42:42,573
A day.

1523
01:42:42,573 --> 01:42:43,115
A hundred a day, which

1524
01:42:43,115 --> 01:42:43,907
is pretty substantial.

1525
01:42:43,907 --> 01:42:45,242
Like for someone I don't use.

1526
01:42:45,617 --> 01:42:47,494
What did we get before like 40 a day when

1527
01:42:47,494 --> 01:42:49,454
GPT 4 first came out or something?

1528
01:42:49,454 --> 01:42:50,622
Or 50 a day or whatever it was.

1529
01:42:50,706 --> 01:42:51,790
I don't think I would use that many.

1530
01:42:51,790 --> 01:42:53,709
So I think that's, I'm pretty excited.

1531
01:42:53,709 --> 01:42:55,127
And I'm unfortunately, as I said, it's

1532
01:42:55,127 --> 01:42:56,879
not available here in Australia.

1533
01:42:56,920 --> 01:42:59,214
Um, but yeah, I don't know.

1534
01:42:59,214 --> 01:43:00,674
I just, oh wait, that's

1535
01:43:00,674 --> 01:43:01,925
no, we're talking operator.

1536
01:43:01,925 --> 01:43:02,634
I'm mixing it up here.

1537
01:43:03,051 --> 01:43:05,596
Sam talking about O3 mini not operator.

1538
01:43:06,138 --> 01:43:08,307
That's what I know.

1539
01:43:08,307 --> 01:43:09,474
He said that they did say something about

1540
01:43:09,474 --> 01:43:11,852
it coming to that operator to plus.

1541
01:43:12,060 --> 01:43:12,561
Yeah.

1542
01:43:12,561 --> 01:43:14,354
They'll bring it as soon as they can.

1543
01:43:14,354 --> 01:43:14,980
It's like, yeah.

1544
01:43:15,731 --> 01:43:16,565
And they didn't announce

1545
01:43:16,565 --> 01:43:17,608
how well the usage was.

1546
01:43:18,233 --> 01:43:18,901
Um, interesting.

1547
01:43:19,610 --> 01:43:21,069
But like just the monitoring the

1548
01:43:21,069 --> 01:43:22,195
detection, I've just got the

1549
01:43:22,195 --> 01:43:23,947
open-air article up with over the late

1550
01:43:23,947 --> 01:43:25,032
road talking about it.

1551
01:43:25,032 --> 01:43:25,282
It's good.

1552
01:43:25,782 --> 01:43:27,576
Um, there, you know, they hint, there's a

1553
01:43:27,576 --> 01:43:28,827
lot of safety and privacy.

1554
01:43:29,202 --> 01:43:31,580
So the, you know, it's not up here.

1555
01:43:31,872 --> 01:43:33,790
Um, so some of the risks associated with

1556
01:43:33,790 --> 01:43:36,835
what is, you know, what is happening

1557
01:43:36,835 --> 01:43:38,003
with the data, there was a lot of

1558
01:43:38,003 --> 01:43:40,339
concerns when Microsoft released a lot of

1559
01:43:40,339 --> 01:43:41,590
their sort of their co-child

1560
01:43:41,590 --> 01:43:43,300
versions of similar things.

1561
01:43:43,675 --> 01:43:44,468
Um, there was questions

1562
01:43:44,468 --> 01:43:45,844
around safety and security.

1563
01:43:45,844 --> 01:43:47,429
So obviously they're really pushing to

1564
01:43:47,429 --> 01:43:48,889
say that it's transparent.

1565
01:43:49,097 --> 01:43:50,599
I love a good, um, you know, there's,

1566
01:43:50,933 --> 01:43:51,892
yeah, exactly right.

1567
01:43:52,768 --> 01:43:53,143
Okay.

1568
01:43:53,143 --> 01:43:53,936
You have to pull up their graph.

1569
01:43:54,436 --> 01:43:55,312
Um, limitations.

1570
01:43:55,896 --> 01:43:57,773
It's currently an early research preview.

1571
01:43:58,148 --> 01:43:59,358
Um, so I was capable of

1572
01:43:59,358 --> 01:44:01,068
handing a wide range of tasks.

1573
01:44:01,068 --> 01:44:02,069
It's still learning and

1574
01:44:02,069 --> 01:44:02,945
so will make mistakes.

1575
01:44:02,945 --> 01:44:04,321
You may ask it to do something and it

1576
01:44:04,321 --> 01:44:05,864
just won't do the right thing.

1577
01:44:05,906 --> 01:44:07,282
That's expected with tools like this.

1578
01:44:07,491 --> 01:44:08,158
Going back to what you

1579
01:44:08,158 --> 01:44:09,159
said before there and stuff.

1580
01:44:09,242 --> 01:44:10,786
Um, and yeah, they do have the safety

1581
01:44:10,786 --> 01:44:12,537
component here folks, but like, you

1582
01:44:12,537 --> 01:44:14,206
know, I think we have such a higher

1583
01:44:14,206 --> 01:44:16,500
expectation of no mistakes from machines.

1584
01:44:16,667 --> 01:44:18,293
And I definitely have that bias.

1585
01:44:19,127 --> 01:44:21,463
Um, you know, I use bolts to build things

1586
01:44:21,463 --> 01:44:25,050
and it will make mistakes, but, uh, you,

1587
01:44:25,050 --> 01:44:27,010
you tell it like, why did you do that?

1588
01:44:28,220 --> 01:44:28,804
You know, like, I

1589
01:44:28,804 --> 01:44:29,596
think we should read it.

1590
01:44:29,763 --> 01:44:32,474
Gosh, I go full Napoleon dynamite.

1591
01:44:32,849 --> 01:44:35,811
Um, you know, ask Pedro, I should call it

1592
01:44:35,811 --> 01:44:36,395
that actually, there

1593
01:44:36,395 --> 01:44:37,396
should be an ask Pedro

1594
01:44:37,646 --> 01:44:39,773
or vote for Pedro, uh, kind of, um,

1595
01:44:39,940 --> 01:44:40,565
anyway, you'll get the

1596
01:44:40,565 --> 01:44:41,566
reference if you're nearly

1597
01:44:41,566 --> 01:44:43,276
40, um, pretty much.

1598
01:44:44,277 --> 01:44:46,571
But yeah, so operator it's, it's looking

1599
01:44:46,571 --> 01:44:48,240
really interesting and stuff out there.

1600
01:44:48,615 --> 01:44:50,409
Uh, definitely worth checking out.

1601
01:44:50,826 --> 01:44:53,662
Um, I'm speaking of, uh, operator and

1602
01:44:53,662 --> 01:44:54,955
agents and the work.

1603
01:44:55,163 --> 01:44:57,708
So, um, on perplexity, their mobile app,

1604
01:44:57,708 --> 01:44:58,834
if you have the pro version, they've

1605
01:44:58,834 --> 01:45:00,836
also just, uh, released

1606
01:45:00,836 --> 01:45:02,671
their perplexity assistant.

1607
01:45:03,213 --> 01:45:04,506
I'll just play this up here.

1608
01:45:04,506 --> 01:45:08,468
Oh, play me the final song

1609
01:45:08,468 --> 01:45:09,511
from the movie interstellar.

1610
01:45:13,306 --> 01:45:16,226
Help me book the best

1611
01:45:16,226 --> 01:45:17,602
Japanese restaurant for a date.

1612
01:45:26,653 --> 01:45:27,529
Just finished reading this.

1613
01:45:27,863 --> 01:45:28,864
What should I read next?

1614
01:45:38,957 --> 01:45:40,625
Summarize this page and send it to John.

1615
01:45:43,295 --> 01:45:43,795
Yeah.

1616
01:45:47,632 --> 01:45:48,383
Remind me two hours

1617
01:45:48,383 --> 01:45:49,342
before the next four years.

1618
01:46:03,190 --> 01:46:03,523
That's crazy.

1619
01:46:03,815 --> 01:46:05,233
So that was just pretty interesting.

1620
01:46:06,276 --> 01:46:07,694
Um, switch back to my desktop.

1621
01:46:08,320 --> 01:46:09,529
I mean, you know, some of those things

1622
01:46:09,529 --> 01:46:12,115
you, you can do it yourself either just

1623
01:46:12,282 --> 01:46:13,658
without AI and it

1624
01:46:13,658 --> 01:46:14,534
takes you time to do it.

1625
01:46:14,534 --> 01:46:15,535
And then with AI and

1626
01:46:15,535 --> 01:46:16,369
it's a little bit less.

1627
01:46:17,037 --> 01:46:18,872
And now like just say,

1628
01:46:19,122 --> 01:46:20,207
and it will be automated.

1629
01:46:20,832 --> 01:46:21,291
It's crazy.

1630
01:46:21,291 --> 01:46:22,334
Look me flights to XYZ.

1631
01:46:22,584 --> 01:46:23,710
Do what do you do with the

1632
01:46:23,710 --> 01:46:25,045
rest of your time kind of thing?

1633
01:46:25,796 --> 01:46:26,922
You know, I mean, it begs

1634
01:46:26,922 --> 01:46:28,381
to remember who came out.

1635
01:46:28,381 --> 01:46:30,342
It, uh, reduced our attention span.

1636
01:46:30,342 --> 01:46:31,927
I don't remember what to find information

1637
01:46:31,927 --> 01:46:34,638
from, you know, we had to go that,

1638
01:46:34,638 --> 01:46:36,598
that one hour wait period was right down

1639
01:46:36,598 --> 01:46:37,849
to like five minutes.

1640
01:46:38,100 --> 01:46:39,476
I would ask information now.

1641
01:46:39,768 --> 01:46:40,769
I would ask Jeeves.

1642
01:46:41,103 --> 01:46:41,561
Yeah. That was my thing

1643
01:46:41,561 --> 01:46:43,021
before PageRank came out.

1644
01:46:43,021 --> 01:46:44,397
Thank you, Google, you know, for that

1645
01:46:44,397 --> 01:46:46,650
innovation of PageRank, but like I,

1646
01:46:47,150 --> 01:46:49,027
you searched manually yourself.

1647
01:46:49,027 --> 01:46:49,236
Yeah.

1648
01:46:49,236 --> 01:46:50,278
Now I don't want to complete

1649
01:46:50,278 --> 01:46:51,613
any tasks if it can't be done.

1650
01:46:52,155 --> 01:46:53,406
That is crazy.

1651
01:46:53,698 --> 01:46:54,449
Can you look some flights?

1652
01:46:54,908 --> 01:46:55,408
Sorry, honey.

1653
01:46:55,408 --> 01:46:56,493
It's going to take me 10 minutes.

1654
01:46:56,493 --> 01:46:57,160
I can't handle it.

1655
01:46:57,160 --> 01:46:57,369
Yeah.

1656
01:46:57,369 --> 01:46:59,788
So I have to call the, uh, the, um, I

1657
01:46:59,788 --> 01:47:00,580
mean, people still go

1658
01:47:00,580 --> 01:47:02,541
to a flight center and,

1659
01:47:02,999 --> 01:47:04,084
you know, people travel agents.

1660
01:47:04,459 --> 01:47:04,709
People do.

1661
01:47:04,918 --> 01:47:05,877
And I think it's useful.

1662
01:47:05,877 --> 01:47:07,587
Like you want that human touch with some

1663
01:47:07,587 --> 01:47:08,213
of these things, right?

1664
01:47:08,213 --> 01:47:09,005
It's also useful because

1665
01:47:09,005 --> 01:47:10,215
if something goes wrong,

1666
01:47:10,674 --> 01:47:11,258
agents take

1667
01:47:11,258 --> 01:47:13,093
responsibility for support and help.

1668
01:47:13,510 --> 01:47:13,969
Whereas if you do it

1669
01:47:13,969 --> 01:47:14,845
yourself, you don't have that.

1670
01:47:15,095 --> 01:47:15,720
That's, that's, that's the

1671
01:47:15,720 --> 01:47:16,972
main reason why you go to a,

1672
01:47:17,597 --> 01:47:18,682
why you go to one of those.

1673
01:47:18,682 --> 01:47:19,266
Humans can think

1674
01:47:19,266 --> 01:47:20,267
outside the box and stuff.

1675
01:47:20,475 --> 01:47:21,309
So yeah, really interesting.

1676
01:47:21,601 --> 01:47:22,853
Until, until complexity came along.

1677
01:47:23,186 --> 01:47:23,603
Exactly.

1678
01:47:23,603 --> 01:47:23,979
I didn't know all

1679
01:47:23,979 --> 01:47:25,021
that stuff on perplexity.

1680
01:47:25,021 --> 01:47:26,439
I mean, there's a whole lot of, uh, and

1681
01:47:26,439 --> 01:47:27,524
sorry, I just see here,

1682
01:47:27,607 --> 01:47:29,109
uh, Matt Wolf, who we watch a lot of

1683
01:47:29,109 --> 01:47:31,069
videos and he's been going crazy with

1684
01:47:31,069 --> 01:47:32,737
like the videos recently, I think he was

1685
01:47:32,737 --> 01:47:34,573
only doing like a weekly kind of thing.

1686
01:47:34,573 --> 01:47:36,825
And then, um, sometimes updates, but now

1687
01:47:36,825 --> 01:47:38,076
it feels like every day there's a few

1688
01:47:38,118 --> 01:47:39,744
things from him just because there's so

1689
01:47:39,744 --> 01:47:42,664
much going on, but he, he just put it all

1690
01:47:42,664 --> 01:47:44,833
out from like Friday, what he saw.

1691
01:47:44,833 --> 01:47:46,126
Like, so you mentioned the deep seek, he

1692
01:47:46,126 --> 01:47:47,586
showed the deep seek integration there.

1693
01:47:47,586 --> 01:47:48,628
We can see, um,

1694
01:47:48,628 --> 01:47:49,838
perplexity, like have that.

1695
01:47:50,255 --> 01:47:51,006
Um, I actually know

1696
01:47:51,006 --> 01:47:51,923
that was like operator.

1697
01:47:52,507 --> 01:47:53,550
God, there's so much stuff.

1698
01:47:53,550 --> 01:47:55,343
It'd be nice if there was like just one

1699
01:47:55,343 --> 01:47:58,180
thing, you know, it's O3 mini or it's

1700
01:47:58,180 --> 01:48:00,182
operator or it's deep sea guys.

1701
01:48:00,515 --> 01:48:00,974
Just, just

1702
01:48:01,224 --> 01:48:02,767
slow down, please slow down.

1703
01:48:02,767 --> 01:48:03,518
There's too much happening.

1704
01:48:03,518 --> 01:48:05,103
I mean, I feel like I think I did watch

1705
01:48:05,103 --> 01:48:06,313
one of Matt Wolf's recent episodes

1706
01:48:06,479 --> 01:48:08,273
and he summed up nicely as like, it feels

1707
01:48:08,273 --> 01:48:10,025
like January has already been an entire

1708
01:48:10,108 --> 01:48:13,278
year of just stuff coming out, but it's

1709
01:48:13,278 --> 01:48:15,363
also talking about how fast everything in

1710
01:48:15,363 --> 01:48:17,824
AI is moving right now with the rise of

1711
01:48:17,824 --> 01:48:19,367
agents with operator,

1712
01:48:19,868 --> 01:48:21,036
um, and other, other

1713
01:48:21,077 --> 01:48:22,913
businesses like having some really

1714
01:48:22,913 --> 01:48:24,164
amazing agent style

1715
01:48:24,164 --> 01:48:26,166
tools, um, through to these

1716
01:48:26,208 --> 01:48:29,419
improved models, um, deeper thinking code

1717
01:48:29,419 --> 01:48:31,630
generators, be crazy way more accurate,

1718
01:48:31,630 --> 01:48:33,423
less hallucinations across other tools.

1719
01:48:33,423 --> 01:48:34,090
It's like.

1720
01:48:34,132 --> 01:48:36,843
You know, there was, there's that saying

1721
01:48:36,843 --> 01:48:38,303
that, um, sometimes,

1722
01:48:38,303 --> 01:48:41,223
uh, months like this, it

1723
01:48:41,223 --> 01:48:43,475
feels like decades happen in a month and

1724
01:48:43,475 --> 01:48:47,145
there's other, um, I don't know.

1725
01:48:47,437 --> 01:48:48,855
Other than what it feels like for like,

1726
01:48:48,855 --> 01:48:50,106
if not, that happens next month.

1727
01:48:50,106 --> 01:48:51,274
If we get to the end of February, we're

1728
01:48:51,274 --> 01:48:53,276
like, nothing's happened since deep sea

1729
01:48:53,276 --> 01:48:53,693
clock.

1730
01:48:53,693 --> 01:48:53,902
Yeah.

1731
01:48:53,902 --> 01:48:55,612
What the slow kind of, yeah.

1732
01:48:55,862 --> 01:48:57,697
Took my, I had my, I had my

1733
01:48:57,697 --> 01:49:00,325
own, uh, fuel to feel by a fuel.

1734
01:49:02,369 --> 01:49:03,495
Crystal never limp that down.

1735
01:49:03,995 --> 01:49:04,621
Until we will.

1736
01:49:04,913 --> 01:49:06,581
Um, what other news have you got?

1737
01:49:06,581 --> 01:49:07,958
Cause I'll have some examples

1738
01:49:07,958 --> 01:49:09,501
to round this out and stuff.

1739
01:49:09,501 --> 01:49:09,793
Yeah.

1740
01:49:09,793 --> 01:49:11,127
So like, uh, just some small things

1741
01:49:11,127 --> 01:49:11,962
that's happened out there.

1742
01:49:11,962 --> 01:49:13,255
We talked about, um, we talked about

1743
01:49:13,255 --> 01:49:14,756
perplexity with their new agent tool.

1744
01:49:15,882 --> 01:49:17,300
Um, Oh, we're sharing on my screen.

1745
01:49:17,926 --> 01:49:18,885
That's a, this is kind

1746
01:49:18,885 --> 01:49:22,430
of, um, put the thing up.

1747
01:49:23,431 --> 01:49:23,640
Yeah.

1748
01:49:25,600 --> 01:49:27,394
Um, so some other small things we talked

1749
01:49:27,394 --> 01:49:29,729
about perplexity with their app, um,

1750
01:49:29,729 --> 01:49:31,898
and the agents that, that you can now

1751
01:49:31,898 --> 01:49:32,899
pretty much do anything on the

1752
01:49:32,941 --> 01:49:34,401
application that has

1753
01:49:34,401 --> 01:49:35,277
access to your phone.

1754
01:49:35,527 --> 01:49:36,820
It's interesting because it's kind of

1755
01:49:36,820 --> 01:49:37,988
like a model that keeps you in the

1756
01:49:37,988 --> 01:49:39,239
application on your phone, almost like a

1757
01:49:39,239 --> 01:49:40,991
little mini OS device.

1758
01:49:41,491 --> 01:49:42,909
Now, if we're tools like this coming out.

1759
01:49:42,909 --> 01:49:43,243
I mean, that's what

1760
01:49:43,243 --> 01:49:44,744
the rabbit wanted to do.

1761
01:49:44,995 --> 01:49:45,161
Exactly.

1762
01:49:45,370 --> 01:49:45,537
Right.

1763
01:49:45,787 --> 01:49:47,163
And other tools out there.

1764
01:49:47,163 --> 01:49:48,623
So it's interesting if all things like

1765
01:49:48,623 --> 01:49:50,542
the rabbit start, you know, becoming.

1766
01:49:51,167 --> 01:49:51,501
Redundant.

1767
01:49:51,918 --> 01:49:54,170
Well, like some other new device will

1768
01:49:54,170 --> 01:49:55,839
come out with a more approved experience

1769
01:49:55,880 --> 01:49:58,508
using, you know, the, uh, the R1 model

1770
01:49:58,508 --> 01:49:59,426
next to a something like

1771
01:49:59,426 --> 01:50:00,593
a Pepsi from an agent perspective.

1772
01:50:01,094 --> 01:50:02,470
It's a wonder whether you actually,

1773
01:50:02,679 --> 01:50:04,014
because the rabbit was like an extra

1774
01:50:04,014 --> 01:50:05,348
kind of device and you still needed your

1775
01:50:05,348 --> 01:50:06,433
phone and stuff like that.

1776
01:50:06,433 --> 01:50:08,643
So the complaint was, well, why, why

1777
01:50:08,643 --> 01:50:10,520
don't we just see this on the phone?

1778
01:50:10,520 --> 01:50:11,813
And that's what we're seeing now.

1779
01:50:11,813 --> 01:50:11,980
Exactly.

1780
01:50:12,188 --> 01:50:12,355
Right.

1781
01:50:12,355 --> 01:50:14,024
Interesting seeing forecasts player.

1782
01:50:15,150 --> 01:50:17,360
Um, yeah, some other small news.

1783
01:50:17,694 --> 01:50:19,988
Google's just invested a billion dollars

1784
01:50:19,988 --> 01:50:22,115
in additional into anthropic, which is

1785
01:50:22,115 --> 01:50:22,532
interesting.

1786
01:50:22,532 --> 01:50:23,992
I didn't realize that Google has such a

1787
01:50:23,992 --> 01:50:25,118
big investment into them.

1788
01:50:25,118 --> 01:50:27,245
And we've already know that, um, Amazon

1789
01:50:27,245 --> 01:50:28,997
has invested multiple millions of

1790
01:50:29,331 --> 01:50:30,665
dollars into anthropic, one of the

1791
01:50:30,665 --> 01:50:32,208
largest investors, but Google is also

1792
01:50:32,208 --> 01:50:33,168
coming to the field.

1793
01:50:33,918 --> 01:50:34,127
Yeah. Yeah.

1794
01:50:34,669 --> 01:50:37,172
Um, and speaking of investments, so we've

1795
01:50:37,172 --> 01:50:40,008
talked about Stargate over the last, uh,

1796
01:50:40,008 --> 01:50:41,301
Oh, great.

1797
01:50:41,509 --> 01:50:42,218
Great movie.

1798
01:50:42,886 --> 01:50:43,470
Colonel Jackson.

1799
01:50:43,762 --> 01:50:45,722
The movie, I love the movie better than

1800
01:50:45,722 --> 01:50:47,015
the series, to be honest.

1801
01:50:47,015 --> 01:50:49,225
The OG movie is, is a classic and I still

1802
01:50:49,225 --> 01:50:50,435
love it, but I'm a big fan of the series

1803
01:50:50,435 --> 01:50:50,810
as well.

1804
01:50:50,977 --> 01:50:52,729
I absolutely stopped talking.

1805
01:50:52,729 --> 01:50:53,980
I was a Stargate fan forever.

1806
01:50:54,189 --> 01:50:54,814
We can't.

1807
01:50:54,814 --> 01:50:56,941
Um, but joking with our Stargate, they're

1808
01:50:56,941 --> 01:50:57,984
$500 billion investment.

1809
01:50:58,318 --> 01:51:00,195
Uh, apparently soft bank has released

1810
01:51:00,195 --> 01:51:01,654
that first 25 billion.

1811
01:51:02,155 --> 01:51:03,531
Apparently, um, you aren't

1812
01:51:03,531 --> 01:51:04,407
going to question it all.

1813
01:51:04,699 --> 01:51:04,908
Obviously.

1814
01:51:05,033 --> 01:51:05,241
Yes.

1815
01:51:05,492 --> 01:51:07,744
And further investments will be, um,

1816
01:51:07,744 --> 01:51:09,746
getting unleashed as

1817
01:51:09,746 --> 01:51:11,247
they move into making

1818
01:51:11,373 --> 01:51:13,249
Stargate a reality, um, which is

1819
01:51:13,249 --> 01:51:13,750
obviously the

1820
01:51:13,750 --> 01:51:15,043
development of huge amounts of

1821
01:51:15,043 --> 01:51:15,960
infrastructure supporting

1822
01:51:15,960 --> 01:51:18,254
open AI and other AI tools.

1823
01:51:18,505 --> 01:51:20,507
I can't wait to open up portals to other

1824
01:51:20,507 --> 01:51:21,508
dimensions with this,

1825
01:51:21,508 --> 01:51:22,634
uh, Stargate project.

1826
01:51:22,675 --> 01:51:23,093
Yes.

1827
01:51:23,218 --> 01:51:25,887
You know, bringing AI to ancient

1828
01:51:25,887 --> 01:51:27,931
Egyptians in another

1829
01:51:27,931 --> 01:51:29,474
world is going to be really

1830
01:51:29,474 --> 01:51:29,808
interesting.

1831
01:51:30,100 --> 01:51:30,809
So, so we'll see.

1832
01:51:31,184 --> 01:51:31,684
Pretty exciting.

1833
01:51:32,435 --> 01:51:35,355
Um, lastly, there's not an AI news, but

1834
01:51:35,355 --> 01:51:36,815
it was, that was something cause I've got

1835
01:51:36,815 --> 01:51:39,150
a bit of an SEO and I, I got you to other

1836
01:51:39,150 --> 01:51:41,736
SEM background was like Chrome has, I

1837
01:51:41,736 --> 01:51:43,029
didn't have been talking about this for a

1838
01:51:43,029 --> 01:51:44,614
while now it's been coming and there's

1839
01:51:44,614 --> 01:51:45,573
been discussions around businesses

1840
01:51:45,573 --> 01:51:48,034
needing to blockade or protect themselves

1841
01:51:48,451 --> 01:51:51,538
from this pending anti tracking.

1842
01:51:52,288 --> 01:51:55,375
Um, uh, I guess tools that are coming up.

1843
01:51:55,375 --> 01:51:55,750
Oh yeah.

1844
01:51:56,126 --> 01:51:57,877
Uh, so Chrome is now officially launched

1845
01:51:57,877 --> 01:52:00,380
their upgrade with a one click to stop

1846
01:52:00,463 --> 01:52:01,131
being tracked.

1847
01:52:01,464 --> 01:52:03,133
So if you're a business out there and

1848
01:52:03,133 --> 01:52:04,384
you're heavily relying on your cookies

1849
01:52:04,384 --> 01:52:06,761
and tracking, it's time to start thinking

1850
01:52:06,761 --> 01:52:08,763
about alternatives because Chrome will

1851
01:52:08,763 --> 01:52:10,807
now block you with a long early or users

1852
01:52:10,807 --> 01:52:12,016
can do one click blocking.

1853
01:52:12,016 --> 01:52:14,185
I mean, yeah, that's going to be crazy.

1854
01:52:14,185 --> 01:52:16,062
Like, cause all that data and I know

1855
01:52:16,062 --> 01:52:18,106
businesses that have this,

1856
01:52:18,231 --> 01:52:19,941
uh, sort of information and it provides

1857
01:52:19,941 --> 01:52:22,902
like such a valuable stream of insights

1858
01:52:23,111 --> 01:52:25,280
to folks in terms of

1859
01:52:25,280 --> 01:52:26,364
what is user behavior.

1860
01:52:26,990 --> 01:52:28,992
When you don't have that, how do you know

1861
01:52:28,992 --> 01:52:30,034
what the user is doing?

1862
01:52:30,034 --> 01:52:31,161
How can you feed that into

1863
01:52:31,161 --> 01:52:32,203
improving your kind of model?

1864
01:52:32,495 --> 01:52:34,164
Like, um, it's going to be interesting.

1865
01:52:34,164 --> 01:52:35,582
There's going to be definitely, if you're

1866
01:52:35,582 --> 01:52:37,876
a popular tool, new ways of

1867
01:52:37,876 --> 01:52:40,044
you getting information, um, and if

1868
01:52:40,044 --> 01:52:41,045
you're not, then

1869
01:52:41,045 --> 01:52:42,589
yeah, it's going to suck.

1870
01:52:42,839 --> 01:52:43,089
Exactly.

1871
01:52:43,548 --> 01:52:44,507
So what do you got?

1872
01:52:44,966 --> 01:52:46,426
Um, so I was just showing up the article,

1873
01:52:46,426 --> 01:52:47,218
just talking about it.

1874
01:52:47,260 --> 01:52:48,970
Uh, nothing really special beyond what

1875
01:52:48,970 --> 01:52:49,637
I've just said, but

1876
01:52:49,637 --> 01:52:51,097
there's, um, discussions

1877
01:52:51,097 --> 01:52:52,265
in the background, what the implications

1878
01:52:52,265 --> 01:52:55,351
are for within, um, corporations are

1879
01:52:55,351 --> 01:52:57,395
humble, obviously weighing in their,

1880
01:52:58,021 --> 01:52:59,063
their sort of opinion

1881
01:52:59,063 --> 01:53:01,232
on this and, uh, and

1882
01:53:01,232 --> 01:53:04,235
the likes of other, um, specialists in

1883
01:53:04,235 --> 01:53:06,237
the market, thinking as like the irony

1884
01:53:06,237 --> 01:53:08,281
of the whole situation, we love irony.

1885
01:53:08,281 --> 01:53:10,366
You know, given that Google's will chrome

1886
01:53:10,366 --> 01:53:11,868
in itself, Google has been one of the

1887
01:53:11,868 --> 01:53:14,037
leaders in, in tracking and cookies,

1888
01:53:14,329 --> 01:53:15,955
cookies and businesses followed along,

1889
01:53:15,955 --> 01:53:17,415
particularly with the whole S with SEM

1890
01:53:17,415 --> 01:53:18,791
and SEO tools that they release.

1891
01:53:19,459 --> 01:53:20,335
And now it's like a one

1892
01:53:20,335 --> 01:53:22,045
one click prevention of that.

1893
01:53:22,045 --> 01:53:22,921
So it's, it's interesting what it's going

1894
01:53:22,921 --> 01:53:24,047
to have on their own implications.

1895
01:53:24,214 --> 01:53:24,881
Cause their own models.

1896
01:53:25,173 --> 01:53:26,132
There's another solution

1897
01:53:26,132 --> 01:53:26,925
that they've got for it.

1898
01:53:26,925 --> 01:53:28,384
They're not going to kill off, you know,

1899
01:53:28,384 --> 01:53:29,302
one part of their business,

1900
01:53:29,469 --> 01:53:31,638
unless they've got a plan out of it.

1901
01:53:31,638 --> 01:53:33,181
And I think anyone

1902
01:53:33,181 --> 01:53:34,182
would do that and stuff.

1903
01:53:34,557 --> 01:53:36,392
Um, I want to round out by showing a few

1904
01:53:36,392 --> 01:53:37,477
examples of things that we're

1905
01:53:37,477 --> 01:53:39,187
working on that will be coming out.

1906
01:53:39,187 --> 01:53:41,314
So, uh, one thing and

1907
01:53:41,314 --> 01:53:42,398
ignore the name here.

1908
01:53:42,398 --> 01:53:42,774
It's not going to be

1909
01:53:42,774 --> 01:53:43,691
called web three Alliance.

1910
01:53:43,733 --> 01:53:45,318
I am thinking of something else.

1911
01:53:45,318 --> 01:53:47,612
It's just a placeholder.

1912
01:53:47,612 --> 01:53:48,363
Let's call it for now.

1913
01:53:48,738 --> 01:53:51,157
But the idea is that, um, I want to

1914
01:53:51,157 --> 01:53:54,536
centralize and centralize, make it easy.

1915
01:53:54,536 --> 01:53:55,245
But you're not centralized.

1916
01:53:55,453 --> 01:53:56,412
I know we are not centralized.

1917
01:53:56,788 --> 01:53:58,581
Uh, but I'm going to make it easier for

1918
01:53:58,581 --> 01:54:00,917
folks to actually access interesting

1919
01:54:00,917 --> 01:54:02,502
kind of research that's going on, whether

1920
01:54:02,502 --> 01:54:04,754
it's videos or like, um, you know, the,

1921
01:54:04,963 --> 01:54:06,422
the PDF like research

1922
01:54:06,422 --> 01:54:07,549
that folks have out there.

1923
01:54:07,590 --> 01:54:09,592
And so for example, we've got the digital

1924
01:54:09,592 --> 01:54:11,386
economy council of Australia and they've

1925
01:54:11,427 --> 01:54:14,347
got a variety of different things that,

1926
01:54:14,347 --> 01:54:16,891
um, they release in terms of, uh,

1927
01:54:17,100 --> 01:54:18,518
research papers and consultations.

1928
01:54:19,102 --> 01:54:20,436
Um, I'll just go back.

1929
01:54:20,436 --> 01:54:23,064
I've got also Oz DeFi and we've got a few

1930
01:54:23,064 --> 01:54:24,857
interesting things here in terms of

1931
01:54:24,857 --> 01:54:26,734
like the meetup transcripts and stuff.

1932
01:54:26,734 --> 01:54:29,153
And so I can go chat with this and, uh,

1933
01:54:29,153 --> 01:54:30,488
you know, select the document and go,

1934
01:54:30,572 --> 01:54:35,201
uh, give me the key, uh, themes from

1935
01:54:35,201 --> 01:54:38,496
this, uh, meetup, for example, and have

1936
01:54:38,580 --> 01:54:40,623
rag running in the background to actually

1937
01:54:40,623 --> 01:54:42,250
give us, uh, some of the answers so

1938
01:54:42,250 --> 01:54:43,626
that we're never working opportunities,

1939
01:54:43,918 --> 01:54:45,169
resources and tools, community

1940
01:54:45,253 --> 01:54:46,713
engagement, blah, blah, blah.

1941
01:54:47,005 --> 01:54:47,922
Um, and you can do a

1942
01:54:47,922 --> 01:54:48,923
whole lot more with that.

1943
01:54:48,923 --> 01:54:51,134
You can select multiple to talk with all

1944
01:54:51,134 --> 01:54:52,302
of these at the same time.

1945
01:54:52,594 --> 01:54:54,304
So this is being worked on.

1946
01:54:54,304 --> 01:54:55,221
It will be released.

1947
01:54:55,638 --> 01:54:57,682
Um, this is the, the, the web tree side.

1948
01:54:57,682 --> 01:54:59,434
I'm going to have the AI side of things

1949
01:54:59,434 --> 01:55:01,728
as well, but just making it so that

1950
01:55:01,728 --> 01:55:03,646
there's a central spot for people to go,

1951
01:55:03,646 --> 01:55:05,231
Oh, where's, you know, where do I go

1952
01:55:05,231 --> 01:55:06,482
for like research on what Aussie

1953
01:55:06,482 --> 01:55:07,984
companies doing and what they're saying?

1954
01:55:08,359 --> 01:55:10,278
Well, we've got folks like jelly C, which

1955
01:55:10,278 --> 01:55:11,863
is a great asset manager out there.

1956
01:55:11,863 --> 01:55:12,947
We'll do some stuff with them.

1957
01:55:13,448 --> 01:55:15,825
Um, ASIC are important and I know they're

1958
01:55:15,825 --> 01:55:16,951
the corporate watchdog, but they

1959
01:55:16,951 --> 01:55:18,786
release really interesting things.

1960
01:55:18,786 --> 01:55:21,247
And if I had this out last year, um, I

1961
01:55:21,247 --> 01:55:23,791
think it was in December, they had a, uh,

1962
01:55:23,791 --> 01:55:26,794
revision of the, uh, the info 225, which

1963
01:55:26,794 --> 01:55:28,671
is their paper on how they're going

1964
01:55:28,671 --> 01:55:30,214
to look at regulating the space.

1965
01:55:30,757 --> 01:55:32,800
And it's needing to have

1966
01:55:32,800 --> 01:55:33,718
that in a central spot.

1967
01:55:33,718 --> 01:55:35,428
So people can like research it and

1968
01:55:35,428 --> 01:55:37,180
compare it to other documents and stuff.

1969
01:55:37,639 --> 01:55:39,140
Um, that is what's going to be released.

1970
01:55:39,140 --> 01:55:41,476
There'll be an AI side of this and I'll

1971
01:55:41,476 --> 01:55:43,353
make the videos that we do here

1972
01:55:43,436 --> 01:55:45,063
available through this as well.

1973
01:55:45,063 --> 01:55:46,439
So it's just easier for people to look at

1974
01:55:46,439 --> 01:55:48,399
AI and blockchain, the stuff that we're

1975
01:55:48,399 --> 01:55:49,692
doing, but if folks are interested in

1976
01:55:49,692 --> 01:55:51,694
contributing to this, I think that's key.

1977
01:55:52,153 --> 01:55:53,863
And then the last thing is, um, speaking

1978
01:55:53,863 --> 01:55:57,367
of agents, um, I am revamping a lot of

1979
01:55:57,367 --> 01:55:58,951
what psych does, for example.

1980
01:55:59,327 --> 01:56:00,787
So here's an example

1981
01:56:00,787 --> 01:56:02,789
of, um, agents at work.

1982
01:56:02,789 --> 01:56:05,375
So I've loaded up my agents here doing a

1983
01:56:05,375 --> 01:56:06,793
few things where if you get the agents

1984
01:56:06,876 --> 01:56:08,836
to work off each other, let's say I want

1985
01:56:08,836 --> 01:56:10,338
to lock the first few agents.

1986
01:56:10,338 --> 01:56:12,382
Cause I only want the last one, the last

1987
01:56:12,382 --> 01:56:13,549
two to actually run.

1988
01:56:14,050 --> 01:56:15,426
Um, and I can tell it which

1989
01:56:15,426 --> 01:56:16,886
agents it needs to look at.

1990
01:56:16,886 --> 01:56:18,721
I can tell it what documents it may need

1991
01:56:18,721 --> 01:56:20,640
to look at because, you know, this is a

1992
01:56:20,640 --> 01:56:22,308
daisy chain, so having the agents

1993
01:56:22,308 --> 01:56:23,893
actually, um, running

1994
01:56:23,893 --> 01:56:26,562
in, uh, in a certain

1995
01:56:26,646 --> 01:56:28,272
order, I think is going to be the key.

1996
01:56:28,272 --> 01:56:29,565
So you can see this agent

1997
01:56:29,565 --> 01:56:31,192
it skipped the first one.

1998
01:56:31,192 --> 01:56:34,153
So just adding these like little, um,

1999
01:56:34,153 --> 01:56:36,072
let's call them UI and UX kind of

2000
01:56:36,114 --> 01:56:37,865
elements that just make it easier.

2001
01:56:38,366 --> 01:56:40,535
Uh, is just so much cooler to do when you

2002
01:56:40,535 --> 01:56:42,120
have tools like, uh, you know, we, we

2003
01:56:42,120 --> 01:56:43,621
use bolts to help us build.

2004
01:56:43,913 --> 01:56:45,581
Um, and yeah, just really

2005
01:56:45,581 --> 01:56:46,749
interesting stuff out there.

2006
01:56:46,749 --> 01:56:48,543
So there's a lot of creation that we're

2007
01:56:48,543 --> 01:56:50,253
going to be providing to, to folks that

2008
01:56:50,253 --> 01:56:52,046
are watchers of this podcast, you know,

2009
01:56:52,338 --> 01:56:54,382
talk to Chris about like the, you'll be

2010
01:56:54,549 --> 01:56:57,093
showing us more, I'm sure soon about the

2011
01:56:57,093 --> 01:56:58,469
AI stuff that you're working on.

2012
01:56:58,845 --> 01:57:00,304
Um, we'll definitely be doing that.

2013
01:57:00,763 --> 01:57:02,265
And I think we're going to be using a lot

2014
01:57:02,265 --> 01:57:03,558
of each other's like tools and stuff.

2015
01:57:03,558 --> 01:57:05,017
Like this is a great time.

2016
01:57:05,351 --> 01:57:07,562
There is a massive opportunity to hyper

2017
01:57:07,562 --> 01:57:10,106
personalize and build in house the

2018
01:57:10,106 --> 01:57:10,690
things that you

2019
01:57:10,690 --> 01:57:11,733
weren't able to do before.

2020
01:57:11,733 --> 01:57:13,401
And if you don't know how to do it, reach

2021
01:57:13,401 --> 01:57:14,652
out, we'll help you out folks.

2022
01:57:14,986 --> 01:57:17,780
But, um, the news, the views, the rants,

2023
01:57:17,780 --> 01:57:19,490
the raves, the building, I think

2024
01:57:19,490 --> 01:57:20,324
is going to be really key.

2025
01:57:21,242 --> 01:57:24,078
Well said on that note, have a great week

2026
01:57:24,078 --> 01:57:25,246
and we'll see you next time.

2027
01:57:25,496 --> 01:57:26,372
Have a great week.

2028
01:57:26,706 --> 01:57:27,915
Feed the fuel of the fire.

2029
01:57:28,249 --> 01:57:30,793
The fuel from your furnace.

2030
01:57:31,043 --> 01:57:32,170
The fuel is fed.

2031
01:57:32,754 --> 01:57:33,463
See you soon folks.

2032
01:57:33,463 --> 01:57:33,671
Bye.

2033
01:57:33,963 --> 01:57:34,172
Peace.